Sunteți pe pagina 1din 248

University of Nottingham

School of Mechanical, Materials, Manufacturing


Engineering and Management

Finite Element Analysis of Glass Fibre Reinforced


Thermoplastic Composites for
Structural Automotive Components
Martin J Wilson
MEng (Hons)

Thesis submitted to the University of Nottingham


for the degree of Doctor of Philosophy
September 2003

Contents
Abstract

vii

Acknowledgements

viii

Glossary

ix

Nomenclature

xi

Chapter 1 Introduction
1.1

Fibre Reinforced Thermoplastic Composites

1-1

1.2

Application of Thermoplastic Composites to Automotive Structures

1-1

1.3

Modelling Damage Development in Thermoplastic Composites

1-4

1.4

Theme of this Research

1-5

Chapter 2 Literature Review


2.1

Introduction

2-1

2.2

Manufacture and Testing of Commingled Thermoplastic Composites

2-1

2.2.1 Processing Commingled Thermoplastic Composites

2-1

2.2.2 Characterisation and Mechanical Testing of Composites

2-4

Crashworthy Applications of Composite Materials

2-5

2.3.1 Energy Absorption Mechanisms in Composite Materials

2-5

2.3.2 Automotive Applications for Structural Composite Materials

2-7

2.3.3 Composite Materials for Side Impact Protection

2-9

Damage Modelling Techniques

2-13

2.4.1 General Macroscopic Progressive Damage Models

2-14

2.4.2 Chang Model

2-17

2.4.3 Ladeveze Model

2-19

2.4.4 PAM-CRASH Bi-Phase Model

2-22

Assessment of Complex Composite Geometries Using Finite Element

2-23

2.3

2.4

2.5

Damage Models
2.6

Conclusion

2-25

2.7

Tables

2-28

2.8

Figures

2-30

Chapter 3 Experimental Methods


3.1

Introduction

3-1

3.2

Material Form

3-1

3.3

Compression Moulding Commingled Thermoplastic Composites

3-1

3.3.1 Tooling Concept

3-2

Material Model Calibration Data

3-2

3.4.1 0/90 Tensile Test

3-3

3.4.2 0/90 Compressive Test

3-4

3.4.3 +45/-45 Tensile Test

3-4

Coupon Tests for Model Validation

3-4

3.5.1 +45/-45 Compressive Test

3-5

3.5.2 0/90 Tensile Hole in Plate Test

3-5

3.5.3 +45/-45 Tensile Hole in Plate Test

3-6

3.5.4 +45/-45 Compressive Hole in Plate Test

3-6

3.6

Conclusion

3-7

3.7

Tables

3-8

3.8

Figures

3-10

3.4

3.5

Chapter 4 Implicit Finite Element Damage Modelling


4.1

Introduction

4-1

4.2

Analysis Technique

4-2

4.2.1 Finite Element Code

4-2

4.2.2 Material Model

4-3

4.2.3 Element Definition and Formulation

4-3

4.2.4 Section Definition

4-4

4.2.5 Definition of Orthotropy Direction

4-4

Development of the Damage Model

4-5

4.3.1 Tensile Failure Model

4-5

4.3.2 Compressive Failure Model

4-6

4.3.3 Shear Damage and Failure Model

4-7

4.3.4 Implementation

4-7

4.3

ii

4.4

Calibration

4-8

4.4.1 Simulation of Calibration Coupon Tests

4-8

4.4.2 Tensile 0/90 Test Simulation

4-9

4.4.3 Compressive 0/90 Test Simulation

4-10

4.4.4 Tensile +45/-45 Test Simulation

4-10

Validation

4-11

4.5.1 Compressive +45/-45 Test Simulation

4-11

4.5.2 Tensile 0/90 Hole in Plate Test Simulation

4-12

4.5.3 Tensile +45/-45 Hole in Plate Test Simulation

4-13

4.5.4 Compressive +45/-45 Hole in Plate Test Simulation

4-13

Model Sensitivity

4-14

4.6.1 Sensitivity to Analysis Step

4-15

4.6.2 Sensitivity to Shear Non-Linearity Parameter

4-15

4.6.3 Sensitivity to Calibration of Damaged Shear Modulus

4-16

4.6.4 Sensitivity to Element Formulation

4-16

4.6.5 Sensitivity to Mesh Refinement

4-18

4.7

Conclusion

4-19

4.8

Figures

4-21

4.5

4.6

Chapter 5 Explicit Finite Element Modelling


5.1

Introduction

5-1

5.2

Analysis Technique

5-1

5.2.1 Finite Element Code

5-1

5.2.2 Material Model

5-2

5.2.3 Element Definition and Formulation

5-2

5.2.4 Section Definition

5-3

5.2.5 Definition of Orthotropy Direction

5-3

Bi-Phase Damage Model

5-3

5.3.1 Degenerate Bi-Phase Damage Model

5-3

5.3

iii

5.4

Calibration

5-4

5.4.1 Calibration of Composite Material Elastic Constants

5-5

5.4.2 Calibration of Composite Material Damage Parameters

5-5

5.4.3 Simulation of Calibration Coupon Tests

5-6

5.4.4 Tensile 0/90 Test Simulation

5-6

5.4.5 Compressive 0/90 Test Simulation

5-7

5.4.6 Tensile +45/-45 Test Simulation

5-8

Validation

5-8

5.5.1 Compressive +45/-45 Test Simulation

5-8

5.5.2 Tensile 0/90 Hole in Plate Test Simulation

5-9

5.5.3 Tensile +45/-45 Hole in Plate Test Simulation

5-10

5.5.4 Compressive +45/-45 Hole in Plate Test Simulation

5-11

Model Sensitivity

5-11

5.6.1 Sensitivity to Damage Model Parameters

5-12

5.6.2 Sensitivity to Mesh Refinement

5-12

5.7

Conclusion

5-13

5.8

Tables

5-14

5.9

Figures

5-15

5.5

5.6

Chapter 6 Application to an Automotive Demonstrator Component


6.1

Introduction

6-1

6.2

Side Intrusion Beam Test Methods

6-2

6.2.1 Small-Scale 3-Point Flexure Test

6-2

6.2.2 Large-Scale 3-Point Flexure Test

6-3

6.2.3 FMVSS214 Side Intrusion Test

6-3

Phase 1 Demonstrator Component

6-4

6.3.1 Phase 1 Demonstrator Component Design

6-4

6.3.2 Phase 1 Demonstrator Component Manufacture

6-5

6.3.3 Phase 1 Demonstrator Component Small-Scale 3-Point Flexure Test

6-5

6.3.4 Phase 1 Demonstrator Component Large-Scale 3-Point Flexure Test

6-6

6.3.5 Phase 1 Demonstrator Implicit Finite Element Damage Modelling

6-8

6.3.6 Phase 1 Demonstrator Explicit Finite Element Damage Modelling

6-10

6.3

iv

Phase 2 Demonstrator Component

6-12

6.4.1 Phase 2 Demonstrator Component Design

6-12

6.4.2 Phase 2 Demonstrator Component Manufacture

6-13

6.4.3 Phase 2 Demonstrator Component Large-Scale 3-Point Flexure Test

6-14

6.4.4 Phase 2 Demonstrator Implicit Finite Element Damage Modelling

6-16

6.4.5 Phase 2 Demonstrator Explicit Finite Element Damage Modelling

6-18

Vehicle Testing

6-21

6.5.1 Installation of Beam in Target Vehicle

6-21

6.5.2 FMVSS 214 Vehicle Test

6-21

6.6

Conclusion

6-22

6.7

Tables

6-24

6.8

Figures

6-25

6.4

6.5

Chapter 7 Discussion and Conclusions


7.1

Introduction

7-1

7.2

Coupon Test Methods for Calibration and Validation of Finite Element

7-2

Damage Models

7.3

7.2.1 Quantity of Specimens Tested and Scatter of Results

7-2

7.2.2 Strain Measurement Techniques

7-3

7.2.3 Shear Behaviour During Off Axis Tests

7-3

7.2.4 Out of Plane Deformation During Compressive Test

7-4

7.2.5 Damage Development in Hole in Plate Specimens

7-4

Implicit Finite Element Damage Modelling

7-5

7.3.1 The Treatment of Fibre Direction and Shear Damage

7-5

7.3.2 Accuracy of Fibre Direction Damaging Behaviour

7-6

7.3.3 Accuracy of Shear Damaging Behaviour

7-6

7.3.4 Application of the Implicit Damage Model to a Complex Component

7-7

7.3.5 Suitability of the Implicit Finite Element Method for Large

7-8

Displacements

7.4

Explicit Finite Element Damage Modelling

7-9

7.4.1 The Treatment of Fibre Direction and Shear Damage

7-9

7.4.2 Accuracy of Fibre Direction Damaging Behaviour

7-9

7.4.3 Accuracy of Shear Damaging Behaviour

7-10

7.4.4 Sensitivity to Calibration Parameters

7-11

7.4.5 Application of the Explicit Damage Model to a Complex Component

7-11

7.4.6 Applicability of Damage Model to Thermoplastics

7-12

7.4.7 Suitability of the Explicit Finite Element Method for Large

7-13

Displacements
7.5

Application of Thermoplastic Composites to Crashworthy Automotive

7-13

Structures

7.6

7.7

7.5.1 Industrial Processing Techniques for Thermoplastic Composite

7-14

7.5.2 The Effect of Process Variability

7-15

7.5.3 The Need for Part Integration

7-15

Recommendations for Future Work

7-16

7.6.1 Inclusion of Residual Load Carrying Capacity in the Implicit Model

7-16

7.6.2 Implementation of the Implicit Damage Model in an Explicit Code

7-16

7.6.3 Assessment of Alternative Damage Models

7-17

7.6.4 Assessment of Relevance of Delamination as a Damage Mechanism

7-17

7.6.5 Rate Dependency

7-17

7.6.6 Development of a Fully Thermoplastic Door Concept

7-17

Conclusions

7-18

Chapter 8 References

Appendices
Appendix A

ABAQUS/Standard User Defined Field

A-1

Appendix B

Derivation of Shear Damage Formulation

A-3

Appendix C

ABAQUS/Standard Material Cards

A-6

Appendix D

PAM-CRASH Material Cards

A-7

Appendix E

CRACTAC Project Publications

A-8

vi

Abstract
This thesis describes the investigation and development of damage modelling
techniques for woven long glass fibre reinforced polypropylene matrix composites.
The objective of the work was to develop and validate predictive models for the
intralaminar damage behaviour of these materials, with the aim of applying the results
to an industrial demonstrator component.

Two damage modelling methods were investigated. The first, based on ply-level
failure criteria and implemented in an implicit finite element code, was developed and
validated using a range of coupon tests for a balanced weave 60% weight fraction
commingled glass/polypropylene composite. The second method utilised a model
previously implemented in the commercial explicit finite element code, PAMCRASH. This model was calibrated and validated using the same coupon tests as the
first model.

The models were subsequently used to simulate an industrial demonstrator


component, during a two-phase design and development programme. The
demonstrator, an automotive side intrusion beam, was designed and predictively
modelled using the two damage modelling techniques investigated.

Finally, the composite component was compared to a steel side intrusion beam, using
a quasi-static vehicle test to a current legislative standard. This test showed
comparable performance in terms of strength and stiffness for the two beams.

It was concluded that the implicit finite element damage modelling technique can
account for the damage and failure modes observed in a woven glass fibre reinforced
polypropylene composite, but is limited when considering high levels of material nonlinearity and damage development, due to the stability of the implicit finite element
method. It was also concluded that the explicit finite element technique was more
suited to the simulation of damage development in thermoplastic matrix composite
components, although the research showed that the model investigated was limited
when considering shear damaging behaviour in a woven fibre reinforced composite.

vii

Acknowledgements
The author would like to thank the supervisors of this work, Dr N. A. Warrior, Dr R.
Brooks and Professor C. D. Rudd, for their guidance.

The research was jointly funded by the EPSRC, the University of Nottingham and the
CRACTAC project consortium, including: Ford Motor Company, Jaguar Cars, BMW,
Borealis, DOW Automotive, ESI Group, Magna, Mapleline, MIRA, Polynorm
Plastics, Park Hill Textiles, SCL, Saint-Gobain Vetrotex and Symalit.

The School of Mechanical, Materials, Manufacturing Engineering and Management is


thanked for the use of departmental facilities and the support of the technical staff,
including Roger Smith, Paul Johns and Dave Smith is acknowledged.

Thanks also go to the other members of the Polymer Composites Research Group,
particularly Nuno Loureno, Dan Bailey, Chris Curtis, Ed Cooper, Richard Fernie,
Craig Wilks and Carlo Santulli, for their invaluable advice and assistance.

Finally, thank you to my friends and family who have supported and encouraged me
throughout this work, my trusty Ford Capri Calypso for reliably getting me to the
office every day and Sallyanne for her belief in me and her love.

viii

Glossary
1:1 weave

Weave pattern with equal number of reinforcing fibres


in the two principal directions

3D fabric

1:1 weave fabric with 1% z-axis fibres

4:1 weave

Weave pattern with twice the number of reinforcing


fibres in one of the two principal directions in-plane

ABAQUS/Standard

Commercial implicit finite element code

Blank

Piece of preconsolidated commingled composite cut to


size prior to heating and moulding

CAD

Computer aided design

Commingled

Intimately combined polymer and reinforcement fibres

CRACTAC

Crashworthy
Automotive
Components
Using
Thermoplastic Composites - Research Programme

Door cassette

Automotive module used to mount door hardware

FEA

Finite element analysis

FORTRAN

Computer programming language

FV

Field variable used in implicit damage model

GMT

Glass mat thermoplastic

Interlaminar

Between plies

Intralaminar

Within a ply

Isothermal moulding

Moulding process where consolidation occurs in a hot


tool - at or above the matrix melt temperature

Lay-up

Combination of plies stacked to form a laminate

Non-isothermal moulding

Moulding process where consolidation occurs in a cool


tool - below the matrix melt temperature

PAM-CRASH

Commercial explicit finite element code

PEEK

Polyetherether-ketone

Ply

Single layer of a laminate

PP

Polypropylene

ix

Preconsolidated

Commingled fabric that has been partially consolidated


to improve moulding and ease of handling

RTM

Resin transfer moulding

SMC

Sheet moulding compound

StaMax

Glass reinforced polypropylene injection moulding


material

Thermoplastic

Polymer softened by heating and hardened by cooling in


a reversible process

Thermoset

Polymer hardened by irreversible chemical change

Tow

Bundle of fibres

TowFlex

Powder impregnated reinforced thermoplastic composite

Twintex

Commingled glass reinforced thermoplastic composite

UD

Unidirectional

VF

Volume fraction

WF

Weight fraction

Nomenclature
ABAQUS/Standard model parameters
E11

Youngs Modulus in first in-plane principal fibre direction

E22

Youngs Modulus in second in-plane principal fibre direction

Poissons Ratio in-plane

G12

Shear modulus in-plane

G13

Shear modulus through thickness

G23

Shear modulus through thickness

T1U

Ultimate tensile strength in first in-plane principal fibre direction

T2U

Ultimate tensile strength in second in-plane principal fibre direction

C1U

Ultimate compressive strength in first in-plane principal fibre direction

C2U

Ultimate compressive strength in second in-plane principal fibre direction

Shear damage nonlinearity parameter

ds

Shear damage level

12

In-plane shear stress

12

In-plane shear strain

12U

Ultimate in-plane shear strain

E11F

Failed Youngs Modulus in first in-plane principal fibre direction

E22F

Failed Youngs Modulus in second in-plane principal fibre direction

Failed Poissons Ratio in-plane

G12D

Fully damaged shear modulus in-plane

G12F

Failed shear modulus in-plane

PAM-CRASH bi-phase model parameters


E11T

Tensile Youngs Modulus in first in-plane principal fibre direction

E11C

Compressive Youngs Modulus in first in-plane principal fibre direction

E22T

Tensile Youngs Modulus in second in-plane principal fibre direction

E22C

Compressive Youngs Modulus in second in-plane principal fibre direction

E33T

Tensile Youngs Modulus through thickness

E33C

Compressive Youngs Modulus through thickness

12

Poissons Ratio in-plane

xi

13

Poissons Ratio through thickness

23

Poissons Ratio through thickness

G12

Shear modulus in-plane

G13

Shear modulus through thickness

G23

Shear modulus through thickness

iT

Tensile initial damage strain

1T

Tensile intermediate damage strain

uT

Tensile ultimate damage strain

d1T

Tensile intermediate damage magnitude

duT

Tensile ultimate damage magnitude

iC

Compressive initial damage strain

1C

Compressive intermediate damage strain

uC

Compressive ultimate damage strain

d1C

Compressive intermediate damage magnitude

duC

Compressive ultimate damage magnitude

[ ]

Numbers in square parenthesis refer to References in Chapter 8

[ ]

Angles (in degrees) in square parenthesis refer to laminate lay-up

xii

Chapter 1
1.1

Introduction

Fibre Reinforced Thermoplastic Composites

Composites combine high strength fibres and lightweight matrices, creating materials
with high specific properties. Through careful selection of fibre length, material and
architecture and the matrix polymer, it is possible to create an extensive range of
engineering materials.

Applications for these materials are varied and wide ranging, from short fibre
reinforced injection moulded thermoplastics for high volume manufacturing, through
to high performance aligned long fibre composites for more demanding applications.
There is a broad range of commercially available resin systems, fibres and preimpregnated composite materials. Each offers a different level of mechanical
performance, surface finish, recyclability, formability and cost.

Thermoplastic matrix based composites have become popular for large volume
production of components and structures, as they offer a number of advantages over
thermosetting composites. They are tough, can be formed or moulded quickly through
the application of heat, they can be recycled easily and produce very little waste
during manufacture. These factors combine to make them appeal strongly to medium
to high volume automotive manufacturers.

1.2

Application of Thermoplastic Composites to Automotive

Structures
Legislation is constantly demanding improvements to every aspect of new passenger
vehicles. This legislation can be simplified into two key requirements. Firstly, cars
must be more environmentally friendly in terms of both fuel efficiency and
recyclability and secondly, they must be safer, offering more protection to both
passengers and pedestrians in the event of an accident. Composites materials can offer
solutions to both these problems.

Increased efficiency is achievable through weight saving since up to 40% of fuel


consumption can be attributed to inertia due to the mass of the vehicle, particularly

1-1

when looking at the urban test cycle [1]. Significant weight reduction, especially in
the body in white, can be achieved by novel design and the use of composites with
higher specific properties than traditional materials such as steel and aluminium.

Recyclability of thermoplastic matrix based composites is also good, components can


be melted to separate the polymer matrix from fibres or the whole composite can be
chopped to produce pelletised materials suitable for injection moulding.

Safety and in particular crashworthiness can also be significantly improved by the use
of composites. Metallic crash structures absorb energy during an impact primarily
through plastic deformation. Composites have the potential to absorb considerably
larger amounts of energy [2] due to damage modes, including matrix deformation,
delamination, local cracking and crushing. The specific energy absorption of
composites has been well documented and shown to offer substantial performance
improvements.

The final requirement and probably the most important to manufacturers in todays
highly competitive passenger vehicle sector is cost. A recent survey by DuPont
Automotive, of automotive design engineers, showed that 50% rated cost as the
number one challenge when designing a new vehicle [3]. This means that as well as
cheap raw materials, cost effective design and manufacturing routes are also required.

Low cost engineering fibres preimpregnated with bulk thermoplastic matrices lend
themselves to forming routes such as non-isothermal stamping or flow moulding.
These materials offer a relatively cheap raw material combined with the rapid
manufacturing technology required by the high volume automotive sector.

Random long glass fibre reinforced polypropylene Glass Mat Thermoplastic (GMT),
is already used widely by the automotive industry for numerous semi-structural
applications. In the year 2000, 37,000 tonnes of GMT were used in the manufacture
of European automotive components, with a further 19,000 tonnes being used in Asia
and the USA. Current production applications include a number of noise shields and
front end structures, as well as the Mercedes A Class rear hatch and double floor
structure, the Volvo 850 rear seat structure and Volvo truck dashboards [4]. Currently
1-2

though, the properties of polypropylene based GMT products preclude them from
being used in fully crashworthy structures.

To produce automotive crash structures with a polypropylene matrix based composite


an aligned or woven fibre composite is required. One such, commercially available,
material is Twintex, a commingled thermoplastic composite produced by SaintGobain Vetrotex. Twintex is available in a number of forms, including various woven
commingled fabrics and preconsolidated sheets. Although much stronger and stiffer
than GMT, Twintex has only found limited use in the automotive sector. Examples of
current applications are rear load floor structures produced by Nissan, off-road vehicle
skid plates by General Motors and a number of truck load area liners and HGV trailer
panels [5]. These structures are still not fully crashworthy applications and offer little
or no contribution towards collision energy absorption in impact situations; they are
being used simply as tough and lightweight semi-structural and low energy impact
protection components.

A key issue facing engineers trying to use these materials is that the forming process
and geometries that can be created using a woven composite are limited when
compared to a flow material such as GMT, which impacts on the cost effectiveness of
using such fabrics. Often to overcome this problem, woven materials are co-moulded
with flow materials to create a structure with complex geometry and improved
structural performance. An example of a production application is Peugeots 806,
glass reinforced polypropylene, bumper structure, which uses GMT co-moulded with
Twintex to significantly increase the flexural stiffness and strength of the part [5].

To fully exploit the crashworthy potential of aligned fibre thermoplastic composite


materials for high volume automotive applications they first need to be shown to offer
one or more advantages over steel or aluminium. The proven ability to mould
complex shapes using co-moulded GMT should allow engineers to produce highly
integrated structures, with aligned fibre materials providing high levels of energy
absorption in critical areas. Before this type of design becomes a reality though,
aligned fibre thermoplastics must be proven as crashworthy materials and design and
analysis tools must be reliable and available to industry.

1-3

1.3

Modelling Damage Development in Thermoplastic Composites

Of the three cost components of a new part: raw materials, manufacture and cost to
design and test, the latter can be considerable, especially when selecting a new
material for a high volume application [6]. A designer needs to be able to develop a
component that will perform satisfactorily, without the need for expensive iterative
testing programmes.

Car designers can, with some level of confidence, design and fully crash test a new
vehicle with primarily metal crash structures, using analysis software, before a single
component has been produced. What manufacturers require is an ability to predict the
performance of thermoplastic composite structures in the same way. Engineers must
be able to model them as they undergo the large amounts of damage seen in vehicle
crash tests or real life accident situations.

There are two basic approaches to modelling the behaviour of a composite material. A
highly detailed micro model of the matrix and fibre system can be used to predict the
development of microcracks and delaminations as the composite material is deformed
and hence a complete and detailed description of the material at all stages can be
obtained. Alternatively a more global approach can be taken. Instead of trying to
describe the complex behaviour of the material at a microstructural level, a
macroscopic approach to identifying damage can be used.

The advantage of the macroscopic approach is that material models can be developed
and calibrated from simple testing of the composite under certain load cases. These
calibrated models can then be used in simulations of large components to predict
global structural behaviour and performance. If understood and used correctly they
can offer designers a much more computationally economic solution to the problem of
simulating damage within composite structures.

Research into the damage modelling of composite materials was primarily undertaken
to aid the design of high performance aerospace structures. The techniques developed
are now being applied to more varied situations. Often in costly aerospace
applications, combinations of high performance matrix and fibre materials have been

1-4

used to produce stiff and lightweight components, designed with large safety factors
and replaced when minimal damage has been identified. Modelling of these structures
has in the past therefore often only required the use of a failure criterion or a simple
and limited damage model. For crashworthy automotive structures these criteria are
not adequate. Vehicle components during a crash undergo large deformations and can
damage extensively before ultimate failure, often progressively absorbing a large
amount of energy.

A designer therefore needs an accurate and computationally efficient solution to the


problem of damage modelling of structures manufactured from composite materials.

1.4

Theme of this Research

This research is part of the CRACTAC (Crashworthy Automotive Components Using


Thermoplastic Composites) project, see Appendix E for publications. The CRACTAC
project is a jointly funded industrial and academic research initiative investigating the
use of reinforced thermoplastics for crashworthy automotive structures. The focus of
this work, within the framework of the CRACTAC project, is the development and
validation of predictive modelling techniques for the in-plane damage behaviour of
long glass fibre reinforced polypropylene composites, with the aim of applying the
results to an industrial demonstrator component.

A review of current work in the field, presented in Chapter 2, has identified a


considerable amount of research into the analysis of damage and failure in thermoset
matrix composites structures, where damage models have been shown to offer the
ability to predict damage initiation and progression. The application of these
techniques to bulk thermoplastic composites is less well documented and little work
has been published on the use of these models for large or complex composite
structures.

In the present work, initial efforts, detailed in Chapter 4, were focussed on the
application and subsequent further development of a thermoset matrix composite
damage model to thermoplastic matrix composites using the ABAQUS/Standard
implicit finite element code. The second stage of the work, presented in Chapter 5,

1-5

was an investigation of the bi-phase material and damage model available to the
analyst in the PAM-CRASH explicit finite element code. Research included an
investigation of the calibration strategies for these damage models and their
sensitivity to the required input parameters.

Current manufacturing and testing methods for thermoplastic composite materials


were also investigated within the scope of the work. These are presented in Chapter 3
and discussed in terms of both the acquisition of relevant data for material model
calibration and the subsequent validation of the models using in-plane damaging test
specimens.

Finally, in Chapter 6, these approaches to the damage modelling of long glass fibre
reinforced polypropylene matrix composites have been compared and contrasted in
terms of the design and modelling of a crashworthy automotive industrial
demonstrator component. This component, a structural side intrusion protection beam
was designed, manufactured and tested, using the techniques developed during the
earlier phases of the study. The demonstrator study has shown that large deformation
and global fibre direction damage development prediction is possible for glass
reinforced thermoplastic composite materials, where the loading is such that damage
modes are predominantly in plane.

To conclude the programme, the concept of a crashworthy glass reinforced


thermoplastic door module, installed in a target vehicle, was tested and compared to a
current steel door structure. This was used to validate the materials and concept as a
viable alternative to steels, when considering structural performance.

1-6

Chapter 2
2.1

Literature Review

Introduction

This chapter presents a review of literature on the manufacturing, testing, design and
predictive damage modelling of woven glass reinforced polypropylene composites.
The review is therefore presented in four sections, the first covering manufacturing
and testing techniques, the second presenting applications of composite materials to
structural automotive components and the final sections reviewing damage modelling
techniques and implementation strategies for finite element analysis codes and their
application to 3D geometries.

2.2

Manufacture and Testing of Commingled Thermoplastic

Composites
Long fibre reinforced thermoplastic composites, due to their high matrix viscosity,
compared to thermoset matrix composites, initially present a challenge for
manufacturing. It is difficult to flow the polymer material to produce a fully
consolidated composite using moulding techniques optimised for thermosets. For this
reason, a range of partially impregnated material forms have been developed, which
allow rapid processing. These combine the thermoplastic and fibre reinforcement
intimately, prior to the main component manufacture phase. Examples of intimately
combined long fibre reinforced thermoplastics include commingled, co-wrapped and
core spun yarns [7], see Figure 2.1.

2.2.1 Processing Commingled Thermoplastic Composites


Twintex is a commingled glass reinforced polypropylene material, which offers cost
effective processing routes, for low, medium and high volume components. It is
available in various states including yarn, woven fabric and pre-consolidated woven
sheets, see Figure 2.2. In this study, the pre-consolidated woven form of the material
is used.

Woven Twintex can be formed into complex parts using a range of manufacturing
methods, which all include three basic stages. Initially the material must be heated
above the melt temperature of the matrix, pressure is then applied to form the
2-1

component and consolidate the composite and finally the material is cooled. The two
most widely used industrial manufacturing processes are compression moulding and
vacuum consolidation. Of the two, compression moulding allows the shorter cycle
time and therefore is the most applicable to medium and high volume automotive
components. It is for this reason that the compression moulding technique was
selected for this study. Compression moulding can be separated further into two
distinct methods, isothermal and non-isothermal moulding. The variation between the
two is predominantly in the moulding and cooling cycle.

An isothermal moulding process forms the pre-heated material in tooling which is at a


temperature high enough to keep the matrix in a molten state. The formed component
is then slowly cooled in the tool while pressure is continuously applied. Nonisothermal moulding uses cool tooling, maintained at a constant temperature, below
the melt temperature of the matrix. The preheated material is transferred to the tool
and pressure is applied, while the matrix rapidly cools and the part is formed. The
non-isothermal process, due to the rapid cooling of the formed part, reduces cycle
time significantly when compared to isothermal moulding. This reduction in cycle
time and the reduced cost associated with cool tooling make this process more
suitable for automotive components.

Isothermal processing of commingled glass reinforced polypropylene has been


thoroughly investigated by Ye et al [8]. Three processing variables are identified as
critical to composite quality. These variables, pressure, time at pressure and moulding
temperature are related to void content, flexural modulus and transverse tensile
modulus, to identify the minimum values required to achieve a satisfactory moulding.
Results from this work show that optimum mechanical properties are achieved when
void content is below 2%. This requires a holding time of over 18 minutes at a
pressure of 1MPa and a temperature of 185C. The work of Klinkmuller et al [9] [10]
confirms this result, suggesting that acceptable moulding quality is achievable with
temperatures and pressures of 175C and 10 bar (~1MPa) respectively, with relatively
small improvement in composite properties above these levels. This shows that
although the isothermal moulding process produces high quality composites, it
requires cycle times that are too long and tool temperatures that are too high for
medium to high volume automotive structures.
2-2

Wakeman et al [11] have investigated the non-isothermal moulding of commingled


glass/polypropylene Twintex fabric. A design of experiments technique was used to
optimise processing variables including, tool temperature, preheat temperature,
moulding pressure and time at pressure. Flexural modulus, flexural strength and void
content were used to measure composite quality. It was shown that preheat
temperature had the largest effect on the quality of the moulding. Results were
relatively insensitive to moulding pressure and time at pressure, as long as these were
above a level of 15MPa and 40s respectively. Below these cut-off values composite
quality was significantly worse, with a void content of over 50% observed in the
poorest specimens, compared to <0.5% in the highest quality mouldings. From the
results of this study, the authors propose processing conditions for optimised
mechanical

performance

of

flat

plaques

manufactured

from

woven

glass/polypropylene commingled fabric, see Table 2.1.

This study was limited to flat plaque specimens manufactured from commingled
fabric using low cost tooling methods. Wakeman [12] suggests that further
investigation using preconsolidated Twintex and matched metal tooling could lead to
equivalent quality mouldings being produced using a shorter time at pressure.
Expansion of this research to more complex three-dimensional geometries, including
curvature, would lead to identification of suitable moulding parameters for an
industrial process.

Osten et al [13] have also presented the moulding of Twintex combined with glass
mat thermoplastic (GMT). During this study, flat plaque specimens of Twintex were
moulded and shown to have a flexural modulus and strength of 12GPa and 300MPa
respectively. This compares to a maximum strength of 259.9MPa and modulus of
13.2GPa reported by Wakeman [11]. The results achieved by Osten et al used metal
tooling and a preheat temperature of 225C.

Bruer and Neitzel [14] present general issues concerning the quality of non-isothermal
compression moulded glass fibre/polyamide commingled thermoplastic composites.
Control of defects such as fibre damage, part distortion, wrinkling and delamination
are discussed. It is suggested that to control wrinkling during forming a fabric
2-3

clamping device is used to tension the fibres as they are drawn into the tool. The
authors also recommend that consolidation pressure is limited to <50MPa to avoid
fibre damage.

2.2.2 Characterisation and Mechanical Testing of Composites


Non-linear composite material models often require a large quantity of calibration
data when compared to models for metallic materials. This data can include both
elastic material properties and a range of post first-ply failure and damage
characteristics, if the model accounts for this behaviour.

Depending on the treatment of damage and consequent effect on mechanical


properties, the range of data required varies significantly. Models that separate fibre
and matrix behaviour can require characterisation of both phases of the composite,
whereas techniques that evaluate damage on a macro scale may require less rigorous
test programmes for calibration. Models may also need experimentally observed
fitting or coupling parameters to be quantified. The calibration parameters required
for various damage models are discussed in more detail in section 2.4.

Curtis [15] proposes a range of tests for full characterisation of a composite material,
for calibration of a homogenous damaging material model, including throughthickness and interlaminar behaviour. Problems were encountered when investigating
through-thickness properties due to the thickness of the moulding required for
specimen manufacture. This is a particular issue for a non-isothermal compression
moulded thermoplastic composite where maximum thickness is limited by the
manufacturing process. Loureno [16] has also shown the proposed test methods to
yield suitable data for the full, ply level characterisation of a thermoset matrix
composite material, for non-linear finite element analysis. The material properties and
associated physical tests are detailed in Table 2.2 and corresponding test specimens
are shown in Figure 2.3.

A certain amount of published data is available for Twintex. Saint-Gobain Vetrotex


present a range of mechanical properties, including elastic constants and failure data
for 60% WF glass/polypropylene Twintex [17]. This data was obtained through a
thorough test programme undertaken by the University of Wyoming (2001) and is
2-4

summarised in Table 2.3. The data differs from published information available in
1998 from the manufacturer [18], also presented in Table 2.3. Variation in modulus
and strength are explained by continuous material improvement undertaken during the
past five years, which is particularly apparent when comparing in-plane shear
modulus. It is also noted that the data presented is dependant on the quality of
manufacturing techniques.

When considering the characterisation of shear behaviour, Pieron and Vautrin [19]
suggest that the 45 tensile test method yields results that are comparable with the
Iosipescu method, in terms of both absolute value and scatter. The 45 tensile test also
offers significant advantages in terms of specimen preparation, investment in test rigs
and complexity of method.

2.3

Crashworthy Applications of Composite Materials

Aligned fibre composite materials, in general, offer a combination of high specific


stiffness, strength and energy absorption when compared to metals. This is observed,
not only in exotic materials, but even when considering bulk thermoplastic matrix
materials such as polypropylene reinforced with glass fibres, see Table 2.4
[12][18][20][21][22]. Composites therefore have the potential to replace metallic
crash energy management structures in transport applications, where low mass and
high strength and energy absorption are key economic drivers.

2.3.1 Energy Absorption Mechanisms in Composite Materials


Composites absorb energy through elastic/plastic deformation and a range of damage
mechanisms. These mechanisms have been characterised as fibre debonding, matrix
cracking and fibre failure [23]. An ideal energy absorber exhibits one or more of these
damage characteristics without catastrophic failure and can be achieved through a
combination of careful design and materials selection.

The final failure modes exhibited by a UD composite ply are presented by Hull [23]
and related to longitudinal, transverse and shear loading, see Figure 2.4. These are
applicable to a thermoplastic matrix composite, although Cantwell and Morten [24]
report that the matrix toughness could significantly affect resistance to certain failure

2-5

modes. In their review of impact resistance of composites they state that selection of a
matrix material with a tensile strain to failure of 4% could lead to improved impact
resistance. Svrinivasan et al [25] confirm this through comparative tangential impact
testing of epoxy and PEEK matrix composites. The post impact damage areas of
various composite plates were investigated and two damage propagation modes were
identified. Damage in the thermoset matrix samples was shown to progress through
the composite by inter-ply delamination and in the thermoplastic matrix samples by
localised shear failure.

Jouri and Shortall [26] performed similar tangential impact tests on random glass
reinforced Nylon 6 composite at various temperatures. Scanning electron microscopy
was used to characterise failure in the composite, post-test. Specimens tested at
ambient temperature exhibited various damage mechanisms including fibre fracture,
localised matrix yielding and cracking, fibre pull out and fibre debonding. No global
delamination is reported in any of the specimens tested. Santulli et al [27] report
similar results for falling weight impact specimens manufactured from woven
commingled glass/ polypropylene composite. Infrared thermography showed that for
plaques impacted with between 15J and 45J of energy, damage remained localised to
the zone of contact with the impactor. Micrographs of the damage zones were used to
identify the damage present in the plaques. None of the balanced weave plaques tested
exhibited any delamination. In most cases, through thickness matrix cracking was
observed as the dominant failure mode.

The most widely investigated energy absorbing composite structure is the crush tube.
This type of structure has the potential to absorb large amounts of energy per unit
mass, through a range of damage mechanisms occurring at, or close to, the crush
front. The energy absorbing crushing modes of a composite tube have been generally
characterised as splaying, fragmentation and local buckling. Predominantly, research
in this area has focussed on composites with thermosetting resin systems [28]. Crush
tubes, although potentially offering a very high level of energy absorption, are
sensitive to various factors, which can lead to unstable collapse. A tube crushed at an
angle can fail catastrophically and only absorb a small percentage of its quoted
capability. Small levels of pre-damage can also significantly affect the performance of
composite crush tubes, making it difficult to assess reparability in crash damaged
2-6

vehicles. Consequently, composite crush tubes have found only limited application in
frontal crash structures, where steel and aluminium are used extensively.

Flexural deformation of composites under tangential loading also produces damage in


the material and can absorb energy. The damage and failure throughout the material
is, in general, not as widespread as that seen in a progressively crushing tube,
although the modes and overall behaviour are not as dependant on such factors as
loading angle and failure initiator geometry. A composite beam in flexure deforms
elastically and then progressively damages. The stability of such components depends
primarily on geometry and material and therefore, for structural components, allows
control over the performance under a range of load cases.

2.3.2 Automotive Applications for Structural Composite Materials


Traditional barriers to composite materials entering high volume passenger vehicle
design, such as processing time, cost and recyclability are being overcome by
thermoplastic matrix composites.

Glass reinforced polypropylene GMT is used extensively in the Mercedes A Class,


where the rear hatch is manufactured as a module, produced and delivered by an
external supplier. This module saves 3kg over a conventional pressed steel structure,
which equates to approximately 25% of the component mass [2]. The BMW Mini also
uses thermoplastic composites in a semi-structural application [29]. In this vehicle a
complete front-end carrier component is manufactured using StaMax, an injection
moulded glass-reinforced polypropylene material. This structure requires 40 fewer
parts when compared to a steel front-end carrier [30], and offers a cost and weight
advantage. This material is also used in front-end applications on the Porsche
Cayenne and the Volkswagen Toureg [31] and in the door module of the Ford Fiesta
[32].

Long glass fibre reinforced polypropylene Twintex has been used for a bumper beam
structure in the Peugeot 806/Evasion van, see Figure 2.5, where the beam has been
shown to perform effectively in both low and high speed collisions [1][33], remaining
in one piece after testing and overcoming the problems of catastrophic collapse and
failure associated with thermosetting matrix composites. Twintex has also been used,
2-7

by Nissan, for the rear load floor of the Primera Break [33], see Figure 2.6. In this
component, aligned fibre reinforced composite facings are moulded as a sandwich
structure with polypropylene honeycomb and polypropylene trim fabric. This use of
100% polypropylene and glass fibre allows relatively easy recycling of the complete
part.

TowFlex, a long fibre, glass reinforced Nylon 6 composite is used in a crashworthy


application on the current BMW M3 [34]. The front and rear bumper beam structures
are thermoformed from preconsolidated sheets of thermoplastic powder impregnated
glass fibres. The bumper crush tubes are also manufactured from the same material,
produced using a continuous compression moulding process. The bumper system,
when tested, was shown to offer improved crash performance, with a weight saving of
60%, over a comparable metallic bumper system [35].

The Lotus Elise uses Resin Transfer Moulding (RTM) to produce structural and
crashworthy components from glass fibre reinforced thermoset resin composites.
These parts include a crush cone structure specifically designed for crash energy
management [1].

Probably the most recent, high profile, use of composites in the automotive industry
has been the Aston Martin Vanquish. This vehicle, launched in 2001 uses carbon fibre
reinforced composites in the crash energy management structures, A-pillars and
transmission tunnel. These parts, not only provide increased stiffness and enhanced
levels of crashworthiness, but offer significant weight saving over similarly
performing steel structures [36].

As well as production parts, research and development work is ongoing, to develop


new composite components for the automotive industry. Some published applications
include structural instrument panels and cross car beams developed by Delphi, Beyer
and General Motors Inc. [37][38][39], a glass fibre reinforced B-post developed in
conjunction with Volvo [40] and an energy absorbing knee bolster designed and
tested by GE Plastics [41]. Much of the current industrial research work remains
unpublished though, due to its commercially sensitive nature [42].

2-8

Two significant barriers to the use of composites in automotive structural design still
remain. The first of these is the End of Life Vehicle Directive [43], which specifies
requirements for recyclability. The second is the relatively high component cost of
composite parts when compared to metals. Aligned fibre reinforced thermoplastic
composites therefore seem a promising option for high volume automotive
applications, if they can be proven to meet both structural and crashworthiness
requirements.

2.3.3 Composite Materials for Side Impact Protection


A large part of this research work was the design and development of a thermoplastic
composite side intrusion protection system. A review of current crashworthiness
requirements and solutions was performed, with a particular focus on the use of
composites in this area.

The two main methods for side impact performance evaluation are the Euro NCAP
test [44] and the Federal Motor Vehicle Safety Standard (FMVSS214) [45]. The Euro
NCAP test uses a deformable barrier mounted on a sled, which is impacted into the
test vehicle. Data is acquired from a fully instrumented dummy mounted in the
driving seat, which is used to evaluate the injury severity of the impact. The
FMVSS214 test has a similar barrier impact, but also includes a simpler quasi-static
pole intrusion. This test evaluates the stiffness of the door and has a pass/fail criteria
based on the resistive force against displacement.

Both tests have been shown to offer an improvement in performance in real life side
impact collision scenarios. A side impact collision involving a car receiving a high
ranking Euro NCAP score, of 4 stars, is 30% less likely to result in fatal or serious
injuries [46]. Similarly, after the introduction of FMVSS214 it was shown that cars
complying to the standard showed a 25% reduction in the risk of serious casualty, in
both vehicle to stationary object and vehicle to vehicle side impact collisions [47].
Further research has shown that although developments are ongoing in the field of
side impact protection, there is still potential for improvement in performance, both in
collisions with fixed roadside objects [48] and with other vehicles [49][50]. In
collisions with other vehicles, pelvic fracture is seen in 85% of cases [49], suggesting
that improved performance of the door and side impact structure could potentially
2-9

reduce the severity and occurrence of the types of injuries seen in the majority of
vehicle to vehicle impacts. Research in Japan, measuring the injury and fatality rate in
all collisions registered by Japanese law enforcement, between 1992 and 1995, has
shown that the side impact fatal injury rate in all reported collisions is 0.32%, which
is considerably higher than the 0.24% fatality rate observed in frontal impacts [51]. It
has also been observed that design of the interior panel of a passenger vehicle door
and the structural collapse mechanism of the B-pillar during side impact can influence
the level of abdominal and thoracic injury [52].

Current production passenger vehicles traditionally use either a pressed or tubular,


high strength steel beam structure for side impact protection. Work has been
undertaken to modularise steel doors [53] and concepts have been shown to offer
weight savings of up to 30% for a full door structure, validated to FMVSS214. These
weight savings were achieved through selection of stronger materials than are
currently used in automotive door structures and through novel design, moving the
side intrusion protection beam to the lower and upper section of the door frame.

Composite materials, as an alternative to steel, have also been investigated, with


varying levels of success. Cheon et al [54] manufactured and tested various side
intrusion protection beam structures from glass fibre-epoxy composite. A square tube,
circular tube and I-beam were tested quasi-statically in three point flexure. Load was
applied to the beams supported over a 470mm span, using a 12 inch diameter
impactor. Results from static tests showed localised catastrophic failure at the point of
load application, between 25mm and 50mm displacement, in all the beams tested. The
square section beam performed better than the other two geometries tested, giving a
peak load of 25.3kN at approximately 30mm displacement. This is similar to the peak
load of 27.3kN observed when testing a steel side intrusion beam, although the steel
beam yielded and failed progressively, resisting load up to the maximum test
displacement of 100mm.

With a catastrophic failure at less than 50mm displacement it is unlikely though that a
beam of this geometry and material combination could meet even the basic quasistatic side intrusion requirement of the Federal Motor Vehicle Safety Standard.

2-10

A hexagonal cross section, glass fibre reinforced epoxy, beam design has also been
proposed by Kamil and Saunders [55] as part of an undergraduate research study. The
beam has been analysed using a linear static finite element technique and assessed
against a maximum stress failure criteria, which showed that the concept could
potentially meet the load displacement performance criteria of FMVSS214. A
prototype beam was not manufactured and this work has not been validated. It is
unlikely that the beam, although a novel design, would meet the structural
requirements of the Federal Standard, since only a basic modelling approach was used
during the design phase.

Patberg et al [56] have investigated an integrated composite door and side impact
protection structure concept. Both thermoset matrix and thermoplastic matrix
composite structures have been designed. The thermoplastic beam was manufactured
from hybrid glass/polypropylene yarns using a combination of braiding and tape
winding, followed by a thermoforming stage. The thermoset composite beam was
manufactured from braided glass fibre reinforced epoxy composite. Basic simulation
of the performance of both concepts has been investigated, although for the
thermoplastic concept it was observed that modelling, using current techniques, was
difficult due to the rigorous material characterisation required. Simulation results are
not presented for either of the concepts. A range of static and dynamic test procedures
has been identified and initial testing of the thermoset concept using a FMVSS214
style pole intrusion has been undertaken. Absolute values are not given, but the
concept appears to perform comparably to a similar steel door structure. The result
does show that the composite concept is initially stiffer than the steel beam and then
undergoes a damage event that reduces the load by approximately 40%, followed by a
steady re-loading. This indicates that a certain amount of catastrophic failure is
occurring in part of the side impact structure. The damage observed in the test is not
discussed so it is difficult to identify whether this failure occurs in the beam or the
door structure.

Twintex glass reinforced polypropylene composite has been used by Erzen et al [57]
to produce a side intrusion beam design that has been investigated using finite
element analysis techniques. The first-ply failure of the composite beam, predicted
using a maximum stress failure criteria, was compared to a steel beam assessed using
2-11

a von Mises yield criteria. Simulation showed that the optimised Twintex beam would
start to fail at approximately 60mm compared to a yield displacement of 84mm for the
reference steel beam. No post failure behaviour modelling or physical testing of
components was undertaken. The study is therefore incomplete in terms of full
validation of either the concept or modelling technique.

A Twintex composite side intrusion beam was also proposed and developed as a
technology demonstrator component for the SACTAC research programme at the
University of Nottingham [58][59], see Figure 2.7. The beam was initially designed
through analysis performed using the ABAQUS/Standard finite element code, with
subsequent simulations being performed using an unvalidated material model for
Twintex implemented in the LS-Dyna explicit finite element analysis code. The final
beam design was manufactured using low cost tooling and tested using a rig
developed for FMVSS214 vehicle testing. The composite beam underperformed by
between 8% and 18% when comparing peak loads and energy absorption with those
of an equivalent steel beam. The major conclusions from this work were that:

Despite the low failure strain of glass/PP composite when compared to steel, a
correctly designed beam could perform as well as a steel beam.

Due regard must be given to the difference in failure stress in tension and
compression for Twintex.

Further work is suggested to improve the design by increasing the dimensions of the
compressive face of the beam to augment the strength. It is also recommended that
simulation techniques should be used to further understand the damage mechanisms
and behaviour of the beam. Development of the component as part of a crashworthy
door module, or inclusion of the beam in a fully thermoplastic door structure, could
also provide significant improvement in performance as well as providing commercial
advantages through parts integration.

Side impact protection is therefore an area where thermoplastic composites have been
shown, by Patberg et al [56] and the work undertaken at Nottingham University
[58][59], to offer the potential to provide a similar level of performance as current

2-12

steel beams. Further work is required including the development and validation of
reliable modelling techniques and the production and testing, to a current standard, of
a composite beam mounted in a vehicle.

2.4

Damage Modelling Techniques

To be able to design composite structures for large deformation, structural and


crashworthy applications, it is important that the behaviour of the material can be
accurately predicted. The basic in-plane damage mechanisms identified for composite
materials and specifically those with thermoplastic matrices have been introduced
previously. In this section of the review, some methods for quantifying this damage
and relating it to degradation in material properties are described. These models will
be discussed with particular focus on their implementation into finite element analysis
codes.

Priston [60] suggests that approaches towards the assessment and quantification of
damage in composite materials can be separated into two distinct concepts. These are,
micromechanical models, which relate individual stresses to distinct microstructural
damage mechanisms and macroscopic models, which describe the damage in a
representative volume of composite. Both these methods can then be used to predict
changes in properties and behaviour of the damaged composite.

The use of micromechanical models as a basis for finite element simulation of


composite materials is computationally expensive, since meshes need to be defined
with a resolution high enough to capture individual damage mechanisms. This would
lead to impractically high mesh density if the models were to be applied to large test
specimens or more complex components. When discussing the use of damage models
with the finite element method, for simulation at component level, the scope is
therefore, in general, constrained to macroscopic approaches.

The volume of literature in this field is extensive and is represented here by a range of
modelling techniques which have been applied to in-plane damage development.
Three of these modelling approaches are discussed in further detail and in section 2.5

2-13

literature regarding the application of damage models to more complex geometries is


presented.

2.4.1 General Macroscopic Progressive Damage Models


The fundamental differences between the majority of published macroscopic damage
modelling techniques are the methods used to identify damage and the subsequent
relationship between the damage identified and the elastic properties of the composite.

Various authors have selected and used failure criteria as the basis of their damage
model, developing a subsequent material property degradation regime based on the
damage mode identified by the criteria. Cheikh [61] presents this type of model in its
simplest form. Maximum longitudinal, transverse and shear stress criteria are used to
identify three distinct damage modes in the composite. These damages, when
identified, lead to a step reduction in the corresponding modulus to zero. The model
proposed is used to simulate progressive damage in a unidirectional composite tensile
test, with various ply orientations, using a stepped analysis, where load is incremented
and damage is computed over a range of small displacements. The methodology
appears to work successfully, although no experimental results are presented, so it is
impossible to validate either the damage identification or modulus reduction methods.
The maximum stress criteria could potentially offer an effective method of identifying
fibre direction damage, where behaviour is dominated by a linear elastic then
fracturing fibre behaviour. It is unlikely that the maximum shear stress criteria could
accurately model shear damage, which is usually a progressive type of damage,
dependant on the matrix material.

A similar method is presented by Belingardi et al [62] who propose a modified Hashin


criteria to describe the failure surface for a composite, with identified failure leading
to a strain-softening behaviour. This model is used to simulate impact damage during
a drop dart test on an E-glass/epoxy plate. In the case of both a 5J and 50J impact,
agreement between experiment and simulation is good. This is in part due to the
introduction of a time dependant behaviour and presumably a calibration of strainsoftening based on experimental observations. Feng et al [63] and Gamble et al [64]
present similar failure criteria based models although the modulus degradation
regimes associated with identified damage, in both cases, do not include a strain2-14

softening behaviour. Both models use criteria to identify three distinct failure modes,
these are: matrix cracking, fibre matrix shear and fibre breakage. Experimental
validation of the Feng model for a composite plate under a uniform pressure loading,
shows good agreement between experiment and simulation, except in the case of
strain levels close to the edge of the plate, where the simulation of boundary
conditions lead to uncharacteristically high strains due to over-constrained edges. The
Gamble model is validated with a hole in plate specimen, where the location and
magnitude of predicted damage is shown to compare well to experimentally observed
results. The authors have suggested that ongoing work to develop and implement a
more physically based, strain dependant damage mechanics model, will improve the
performance of the technique.

Further models of this type presented by Tan and Perez [65], Padhi et al [66] and
Ochoa and Engblom [67] have been validated using a hole in plate under
compression, a plate loaded with a uniform tangential pressure and a plate under four
point bending, respectively. All of these models implement failure criteria in an
incremented finite element analysis, with elastic property degradation based on the
identified damage mode. Chang and Chang [68] also present a failure criteria based
damage model, which is discussed in section 2.4.2.

Chow and Yang [69] describe a more complex damage law based on the deviation
from elastic behaviour, due to a damaged component of the strain, associated with
microcracks and voids in the composite. Similar approaches, described as continuum
damage mechanics models, are also used by Williams et al [70] to simulate crack
propagation in a notched plate and by Vang et al [71] to successfully model the
behaviour of a braided carbon composite tube under cycled pressure loading.
Comparison is made by Vang, between predicted and experimental hoop and axial
strain and shows good agreement for the model. Oytana [72] also presents a damage
mechanics model based on a recoverable, but non-reversible phenomenon treated
separately to damage, which produces an associated permanent plastic deformation.
Results from simulations using this model are compared to experimental results for
composite plates in tension. In the experiment, damage in translucent plates is
measured using an optical technique and is shown to be similar in location and
magnitude to the predicted damage from the model.
2-15

Coats and Harris [73][74] have also undertaken research in the field of continuum
damage mechanics and proposed a volume strain based damage accumulation model.
Experimental validation of the model for a graphite/epoxy composite plate under
tensile loading is presented [75], comparing numerically predicted residual strength
with experimentally obtained results. The model, in this case is within 10% of the
experimental results for a range of notch lengths between 2mm and 23mm.

Iannucci et al [76][77][78] have presented a damage mechanics approach, using


variables which relate to the dissipated energy associated with certain damage modes,
including fibre fracture and fibre-matrix deterioration. These variables are then related
to material characteristics, to represent the reduction in load carrying capacity of a
composite ply associated with accumulated damage. The authors suggest that this type
of approach, implemented in the LS-Dyna explicit finite element analysis code, is a
significantly improved method for predicting impact damage, than traditional stress
based failure criteria modelling techniques [68][79]. Good agreement between
experiment and test is seen for a tangential impact on a woven carbon composite plate
and for a simulation of bird strike on the leading edge of a composite aircraft tail
component. Simulations of the tangential impact using a stress based damage model
shows significant overprediction of laminate strength.

Both the Ladeveze model [80] and the PAM-CRASH bi-phase model [81], discussed
in subsequent sections, are further examples of damage mechanics models, which
have been successfully implemented in commercial finite element codes.

In general, failure criteria based models, which result in material property degradation
based on identified damage modes, can be calibrated using absolute values obtained
through physical test, for example, the tensile failure strength of a ply, or the
compressive strength of the matrix. In contrast, the damage mechanics models require
the user to identify the relation between a theoretical damage level in a volume of
composite and the associated effect on the behaviour in terms of elastic constants, as
well as the plastic behaviour caused by a level of non recoverable deformation due to
cracking in the composite.

2-16

2.4.2 Chang Model


Chang and Chang [68] presented a failure criteria based progressive damage model
for laminated composite plates. The model identifies critical in-plane damage modes
and relates these to reduction in material properties in the damaged areas. The three
in-plane failure modes, describing the failure envelope of the composite, proposed for
a fibre dominated material are: matrix cracking, fibre-matrix shearing and fibre
breakage. The matrix cracking criteria is developed by the authors, based on the
transverse tensile and shear stress in a ply. The fibre/matrix shear damage and fibre
breakage is based on a modified Yamada-Sun failure criterion. The model also
includes a non-linear shear stress-strain relationship based on work by Hahn and Tsai
[82]. Assessment of the performance of the model was made through comparison of
analytical results with experimental results for a carbon/epoxy composite plate, with a
stress concentrator, under tensile loading. Predicted strength and damage showed
good agreement with experiment for various hole diameters.

Subsequently, Chang and Lessard [79] have developed the modelling approach
further, incorporating matrix tensile and compressive failure, fibre buckling and fibrematrix shearing failure. Again, during analysis of composite plates containing a hole
in compression, reduction in material properties based on the type of failure identified
is possible due to the assessment of a range of separate damage mechanisms. Lessard
and Chang [83] have undertaken validation of this model by comparison of
experimental results with the predicted performance of graphite/epoxy composite
plates with various ply orientations. Good agreement is observed between experiment
and simulation for a range of ply angles. The model accurately predicts damage
location and magnitude and the consequent reduction in material performance, as well
as capturing the final failure load of the test specimens.

Further work based on the damage modelling techniques developed by Chang et al


was undertaken by Chang, Liu and Chang [84] to validate the method for tensile
loading on specimens with a stress concentrating hole. The model, previously only
validated for compressive loadings [68][79] was applied to a hole in tensile plate
specimen with various ply orientations. Good agreement was observed between
predicted and observed failure loads and the simulated load/strain behaviour, up to
approximately 1% strain over a 1 inch (25.4mm) gauge length. Since the stress
2-17

concentration is a 0.25 inch diameter hole, this 1% strain over the gauge length
potentially represents up to 4% strain, if deformation is concentrated about the hole.
This simulation of the tensile hole in plate specimen has been repeated by Avalle et al
[85], using a version of the model implemented in LS-Dyna, an explicit finite element
code. The authors state that the models are validated, although no direct comparison
of experimental and analytical results is given.

Shahid and Chang [86] have presented continuing work on accumulative damage
modelling based on damage prediction, using failure criteria developed by Hashin.
Damage, quantified in terms of crack density, is related to composite load carrying
capacity and hence material constants, using ply constitutive equations proposed by
the authors. The model is compared to experimental data for tensile and rail shear
loadings. Stress/strain predictions for various tensile specimens agreed well up to the
point where extensive damage in the specimen resulted in a significant drop in load
carrying capacity. Load and displacement behaviour during the rail shear simulation is
not compared to experimental results, although failure load appears to show good
agreement with experiment. In general the model is accurate when damage is
predominantly in-plane. The authors comment that in laminates which are prone to
delamination, the model may overestimate laminate strength. This is because
delamination damage, which can occur at strain levels significantly below ultimate
failure, results in matrix cracking which can influence global composite strength.

Davila et al [87] have applied the compressive damage model proposed by Chang and
Lessard [79] to the simulation of a ribbed wing-box cover panel manufactured from
graphite reinforced epoxy composite. Experimental results are in good agreement
with the model up to the point at which out of plane deformation occurs in the panel.
Since the model does not include a delamination damage mode, this result was
expected. It is noted that the load displacement response of the panel, up to the point
where out of plane damage occurs, is fairly linear. The model is therefore not fully
validated for highly non-linear behaviour in large components.

In all the work presented by Chang et al [68][79][83], finite element analysis is


undertaken using software tools developed for the purpose of composite plate
analysis. Subsequently, the model proposed by Chang and Lessard [79] has been
2-18

presented as a worked example, input as a user defined material model in the


commercial finite element analysis code, ABAQUS/Standard [88]. The example
analysis presented by the publishers of the code, shows simulation of a carbon/epoxy
composite plate, with a stress concentration, under compressive loading. Experimental
and numerical load displacement curves for this test show excellent agreement up to
1% strain in the 25.4mm (1 inch) gauge length, although above this level, the model is
less able to accurately simulate material behaviour. Experimentally observed ultimate
failure occurs at 2.8% strain with an applied load of 13440N (3000lbs), compared to
2.4% and 11200N (2500lbs) seen in the most accurate of the numerical results. In the
strain range over 1%, the model appears to underestimate the load carrying capacity
of the damaged composite. It is therefore overpredicting either the level of damage in
the composite or the reduction in material properties related to the level of damage
identified. In either case, the result is that the model appears to become less stable,
with larger displacements occurring during the final steps of the analysis, compared to
the start.

In general, these models give a good agreement with experiment, when considering
in-plane damage development in a brittle matrix composite, such as carbon reinforced
epoxy. They accurately capture both fibre direction and shear damage especially
during the initial stages of damage development, although correlation with experiment
appears to reduce at higher strains. There is also little evidence in the literature that
these models have been applied to other materials, especially bulk thermoplastic
matrix composites, such as the one under consideration in this study.

2.4.3 Ladeveze Model


Ladeveze [80][89][90] has proposed a damage mechanics model for laminated
composite materials. This model is based on a damage concept, which relates the
material moduli to parameters describing the damage state of the material, where a
macroscopic damage kinematic is used to quantify the damage parameters. The model
also includes a plasticity coupled to damage, which accounts for the inelastic strains
observed in composites, relating to deterioration of the fibre-matrix interface.

The author proposes that only the shear and transverse tensile moduli vary with
damage state and that the other elastic characteristics remain constant up to a rupture
2-19

point, resulting in the requirement for only two damage parameters for a composite
ply, d the shear damage parameter:

d = 1

G12
G120

(2.1)

E22
0
E22

(2.2)

and d the transverse damage parameter:

d = 1

0
where G120 and E22
denote the undamaged shear moduls and transverse modulus and

G12 and E22 denote the shear moudulus at a damage level d and the transverse
modulus at a damage level d. These parameters combined with the rupture criteria
and plasticity behaviour therefore describe the progressive damaging and failure
behaviour of the composite.

Ladeveze [80] states that this model has been checked on numerous experimental
tests, although only selected results for T300-914 thermoset composite are presented.
It is observed that the model is in good agreement for tensile tests on various fibre
angles, particularly for a tensile test on a [45]2s laminate. Limitations regarding the
final failure or rupture identification when the model is implemented using the finite
element method are also discussed. The post critical behaviour is strongly dependent
on the discretisation of the test specimen, suggesting that convergence studies are
needed to accurately capture cracking damage in the fibre direction.

Ladeveze and Le Dantec [91] present a thorough review of the calibration scheme
required to fully characterise the behaviour of a composite using the model presented
previously. Three tensile tests and a compressive test on various laminates are needed
to derive the parameters used to calibrate the model. These are summarised in Table
2.5. Experimentally these tests do not present a particular challenge although they do
require measurement of transverse laminate strain. It is not possible to apply the
model and calibration strategy, in this form, to a woven fabric reinforced composite.

2-20

This is due to the [67.5]2s, tensile test specimen, which cannot exist when
considering a balanced weave composite.

Allix et al [92][93][94][95] have subsequently utilised the homogenous ply model


proposed by Ladeveze [80], combined with an interface layer to create a mesomodel
of a composite laminate. This approach has been used to successfully model the
intralaminar and interlaminar damage in a carbon fibre/epoxy matrix composite plate
with a stress concentration loaded in tension, when comparing the size and location of
delamination and damage zones [95]. Touchard et al [96] have repeated this study for
an APC-2 carbon fibre/thermoplastic matrix composite. The authors report that when
considering the [45]2s laminate, strain levels in excess of 25% were observed
compared to <10% for the thermoset matrix composite. It is also reported that the
[45]8 specimen for the determination of the transverse damage law only resulted in
calibration up to a damage level of 0.2, since catastrophic failure occurred very
rapidly after damage initiation. The nature of damage observed in the notched tensile
test, showed significant difference in both the type and size of the damage areas when
comparing the thermoset to the thermoplastic matrix composite. No delamination and
only small matrix cracks are observed in the thermoplastic composite, which the
authors conclude, suggests that the model may need some adaptation if it is to be
applied to such materials.

Coutelier and Rozycki [97] have used a version of the model implemented in the
PAM-CRASH finite element code to simulate the behaviour of an E-Glass/Epoxy
composite. The authors conclude that the model is in good agreement when compared
to experiment for a tensile test on a [45]2s laminate and for a dynamic 3 point
bending test on a [90202]s laminate. Results are also discussed for the simulation of a
steel/composite laminated tube crush, although experimental curves are not presented.

Hochard et al [98] have modified the original model present by Ladeveze [68] to
simulate the damaging behaviour of a woven carbon fibre reinforced composite. A
further damage parameter has been included to account for transverse fibre direction
damage. This model, like the work of Coutelier and Rozycki [97] confirms that the
agreement between experimental and simulation results obtained by the original
author [68] are repeatable. Johnson and Simon [99] have also presented preliminary
2-21

work on the implementation of a fabric composite model, based on the unidirectional


model presented previously, in the PAM-CRASH explicit finite element code. The
model has been calibrated for a woven glass fibre reinforced epoxy composite with
verification performed by regenerating the cyclic shear stress/strain curve for the
material. Simulation of a tangential impact on a composite plate is presented, but no
experimental data is shown to validate the result, since this work is part of an ongoing
research programme.

2.4.4 PAM-CRASH Bi-Phase Model


Two material models are available, in the PAM-CRASH finite element code, for the
simulation of damage in composite materials [81]. One of these, the bi-phase model is
described by de Rouvray and Haug [100][101] as an elastic-brittle fibre phase,
superimposed over an elastic-plastic/brittle matrix phase, with a strain based, linear
damaging, ply degradation and failure regime. Pickett et al [102] present an
introduction to the model and detail the implementation in the PAM-CRASH explicit
finite element code. A bi-linear damaging version of then model is also introduced in
this work, see Figure 2.8. Using this version of the model a comparison is made by
Pickett et al [102] between experiment and simulation for impact on a SMC
composite plate, showing good agreement between the test and predicted results.

De Rouvray et al [103] present a numerical investigation of the effect of notches on


the strength of composite plates, using the bi-linear damaging version of the model.
This model is of the same form as that currently available in the commercial analysis
code, PAM-CRASH. Results from this study show that for various ply orientations,
the model can accurately predict ultimate strength. It is noted that the model, in
certain cases, overpredicts the strength of +45/-45 plies, by up to 17%, due to the
lack of an interface layer to capture interlaminar matrix damage. This phenomenon,
also observed by Shahid and Chang [86], reiterates the importance of delamination as
a failure mode in certain laminates. It is expected that this would not have a
significant influence on the performance of a thermoplastic matrix composite though,
for example glass reinforced/PP Twintex, where the tough matrix eliminates
delamination as a critical failure mode.

2-22

In subsequent work, the concept of a modified form of the bi-phase model is also
introduced, where, for cloth or cross-ply laminates the orthotropic constants of the
matrix phase of the model can be used to represent the composite as a whole. Haug
and de Rouvray [104] suggested that this approach is well suited for modelling a
composite laminate using a multi-layered shell element. This use of the modified or
degenerate form of the material model with multi-layered shells has subsequently
been used with some success by Curtis [15] and Lourenco [16].

Haug and Jamjian [105] also propose a programme including 0 and 90 tension, 90
compression and a shear test, as a suitable range of experimental data collection for
the calibration of the model. This is again verified by the work of both Lourenco [16]
and Curtis [15]. It is noted that all of the development, calibration and validation work
presented, concentrates on the use of the model for the simulation of thermosetting
matrix composites. The use of a bulk thermoplastic matrix, such as polypropylene, is
not addressed.

2.5

Assessment of Complex Composite Geometries Using Finite

Element Damage Models


There are predominantly two methods used to solve finite element models, the
implicit and the explicit technique. Due to the nature of these numerical methods, the
implicit and explicit techniques are often applied to two different classes of problem
[122].

An implicit solution is calculated by solving a global stiffness matrix for an


equilibrium loading, to give nodal displacements, and is therefore generally used for
linear or static problems. For problems involving non-linear material behaviour and
dynamic loading, the stiffness matrix is not constant, leading to the need for iterative
schemes to converge a solution for a particular loading. The implicit technique can
therefore become computationally expensive in such situations. Recently, the explicit
technique has become popular for non-linear problems, especially in the field of
crashworthiness, where contact and dynamic effects are critical to the solution.

2-23

An explicit solution, unlike the implicit equilibrium solution, treats the problem as a
dynamic event, considering the equations of motion for nodal displacement, solved in
the time domain. Nodal velocities and displacements, in this case, are obtained by
integrating Newtons second law over a small time-step on an element by element
basis. This leads to a set of uncoupled equations, unlike the implicit technique which
requires assembly and inversion of the complete stiffness matrix.

In general, the explicit technique is therefore selected when significantly high levels
of material nonlinearity and deformation are expected, although the implicit technique
can be used if dynamic effects are negligible and a solution is required for a relatively
small displacement.

Predominantly, the majority of literature in the field of composite materials and


damage modelling focuses on the development and validation of numerical techniques
for the simulation of test coupons or components with simple geometries, using
either the implicit or explicit finite element technique.

In the work presented by Chang et al [68][79][83] the damage modelling techniques


developed are applied only to plates with stress concentrations. There is little
evidence in the current literature that this type of model, implemented in an implicit
finite element code, has been used to simulate more complex geometries. Conversely,
the bi-phase model, implemented in the PAM-CRASH explicit code [81], has been
used for various industrial component failure studies. These include both automotive
and aerospace structures manufactured from composite materials.

Haug and de Rouvray [104], Haug and Jamjian [105][106], Haug et al [107] and
Nakada and Haug [108] present simulations using the model, validated against
experimental data. These include impact on composite plates, composite tube crush
and the simulation and test of a prototype composite car in both side and frontal
impact. The car, manufactured from carbon/epoxy and aramid/epoxy composite
shows that simulation of the response of thermoset matrix composites can yield
realistic load and displacement characteristics when compared to test. Good
agreement is observed between the experimental result of a dynamic side intrusion

2-24

test and simulation, when comparing both the acceleration/time and absorbed
energy/time histories.

Significant further work has been ongoing regarding the application of the bi-phase
model to the simulation of composite materials in the aerospace industry. The concept
of explicit codes being applied to such components was discussed by Johnson et al
[109]. Investigation and calibration of the bi-phase model for these applications was
presented by Kohlgruber and Kamoulakos [110]. This work suggests that for the two
types of fabric composite in question, carbon/epoxy and aramid/epoxy, the degenerate
form of the bi-phase model is most suitable. This is confirmed by the research of
Deletombe et al [111] and McCarthy et al [112] who present the results from
simulation and test of a composite helicopter sub-floor component crush. Agreement
between experimental results and those predicted by the degenerate bi-phase model in
PAM-CRASH are generally acceptable, although some limitations with the model are
suggested. Future research, developing material models specifically for fabric
composites and the consideration of strain rate effects and delamination, is ongoing
by the authors.

Recently, McCarthy and Wiggenraad [113] further reinforced previous findings,


drawing various conclusions, including that the degenerate bi-phase material model
did not allow matching of off-axis behaviour satisfactorily for the materials under
investigation and that a new fabric model, under development at the time of
publication could provide better results in future. This work also addressed the issue
of the need for a residual strength characteristic to be included past the failure
observed in coupon tests, to allow correct prediction of energy absorption. It is
suggested that calibration of post ultimate failure behaviour, a non-physical material
characteristic included to maintain stability during dynamic analysis, could be
achieved through sub-component testing and modelling.

2-25

2.6

Conclusion

This chapter has presented an introduction to Twintex, a woven glass fibre reinforced
polypropylene matrix composite material, summarising current applications and
material property information available in the public domain. The non-isothermal
compression moulding technique has been introduced and a processing window has
been defined based on work at the University of Nottingham by Wakeman et al
[11][12]. General testing of composites for characterisation and specifically the
mechanical performance of Twintex has also been discussed. Furthermore, a range of
tests has been identified as suitable for acquisition of the material properties required
in this study.

A review of current structural applications of composite materials in the automotive


industry has also been performed. Areas where composites potentially offer an
alternative to metallic structures have been discussed. Side intrusion protection has
been identified as a suitable application for the candidate material in this study.
Limited work has been undertaken in this area, although initial results appear to
suggest that there is the potential for a high performance composite side intrusion
protection structure to offer a viable alternative to current metallic design solutions.

The final area of the review has been the application of predictive damage models and
finite element analysis to composites. A range of techniques have been developed and
validated, predominantly for simple geometries, although some research has applied
these methods to more complex structures. There is very little literature available
investigating the use of these techniques for predicting accumulated damage in bulk
thermoplastic matrix composites, such as glass reinforced polypropylene, the
candidate material in this study.

The following three areas have therefore been identified as requiring further
investigation:

1. The application of failure criteria based damage modelling techniques to


woven glass fibre reinforced polypropylene matrix composites.

2-26

2. The use of the bi-phase material model, implemented in the PAM-CRASH


explicit finite element analysis code, for the simulation of the behaviour of
woven glass fibre reinforced polypropylene matrix composites.

3. The use of glass fibre reinforced polypropylene composite materials for side
impact protection and the application of finite element damage modelling
techniques to the design of such components.

2-27

2.7

Tables
Processing Parameter
Tool temperature
Commingled fabric preheat temperature
Moulding pressure
Consolidation time at pressure

Optimised Value
60C
220C
40 bar
80s

Table 2.1
Optimised processing parameters for non-isothermal compression
moulding of commingled glass polypropylene composites [11]

Required Properties
E1t, 12, 13
E2t, 21, 23
E3t, 31, 32
E1c
E2c
E3c
G12
G23
G31
Table 2.2

Evaluating Test
In-plane 0o Tensile Test
In-plane 90o Tensile Test
Through-Thickness Tensile Test
In-plane 0o Compressive Test
In-plane 90o Compressive Test
Through-Thickness Compressive Test
Iosipescu Shear Test
Iosipescu Shear Test
Iosipescu Shear Test

Material properties and associated physical tests [16]

Material Constants
E11
E22
E33
12
21
G12
G13
G23

Published 2003
13.79 GPa
12.97 GPa
- GPa
0.10
0.12
1.72 GPa
1.79 GPa
1.66 GPa

Published 1998
13.6 GPa
13.6 GPa
5.3 GPa
0.08
0.08
1.20 GPa
1.52 GPa
1.52 GPa

Strength

Published 2003
287.6 MPa
265.9 MPa
154.5 MPa
151.1 MPa
18.8 MPa
13.7 MPa
12.1 MPa

Published 1998
313 MPa
313 MPa
125 MPa
125 MPa
25 MPa
31 MPa
31 MPa

T1U
T2U
C1U
C2U
12U
13U
23U
Table 2.3

60% WF balanced weave Twintex properties, 2003[17] and 1998 [18]

2-28

Material

WF

SMC Chopped Glass


Random Glass/Vinyl-ester
GMT Random Glass/PP
Twintex 1:1 Glass/PP (2003)
Twintex 1:1 Glass/PP (1998)
Twintex 1:1 Glass/PP (1998)
Twintex 4:1 Glass/PP (1998)
Twintex UD Glass/PP
Plytron UD Glass/PP
Commingled 1:1 Carbon/PEEK
Commingled UD Carbon/PEEK
Aluminium 6061 T6
Steel SAE 1010
Steel SAE 4340
Table 2.4

(%)
30
38
40
60
60
75
60
75
60
61
61
-

Density
3

(tonne/m )
1.9
1.2
1.5
1.5
1.75
1.5
1.75
1.48
1.6
1.5
2.71
7.87
7.83

TU

specific

(GPa)
9.0
8.6
6.6
13.79
13.6
21
24
38
28
63
145
69
200
200

(MPa)
76
124
120
288
313
420
500
800
650
780
1840
310
365
1034

40.0
100.0
191.7
208.7
240.0
333.3
457.1
439.2
487.5
1226.7
114.4
46.4
132.1

Comparison of the mechanical properties of composite and metallic

materials [12][18][20][21][22]

Test
Tensile test on [0,90]2s laminate
Compressive test on [0,90]4s laminate
Tensile test on [45]2s laminate
Tensile test on [67.5]2s laminate
Table 2.5

Calibration
Fibre tensile strain limit
Fibre compressive strain limit
Damage and plasticity laws
Transverse tension damage master curve

Tests used to calibrate Ladeveze damage model [74]

2-29

2.8

Figures

Figure 2.1
Methods for intimately combining thermoplastic matrix and long fibre
reinforcement [7]

Figure 2.2

Woven commingled Twintex sheet and fabric

2-30

(i)

(ii)

(iii)

(iv)

Figure 2.3
Coupon test specimens: (i) Tensile, (ii)Through-thickness, (iii)
Compressive, (iv) Shear [16]

Figure 2.4

Unidirectional composite failure modes [23]

2-31

Photograph: Peguform, France


Figure 2.5

Peugeot 806/Evasion glass/PP GMT/Twintex bumper beam structure

Photograph: Peguform, France


Figure 2.6

Nissan Primera Break PP Twintex/honeycomb structural load floor

Figure 2.7

SACTAC Project PP Twintex side intrusion beam component

2-32

Figure 2.8

Bi-Phase composite material damage model [81]

2-33

Chapter 3
3.1

Experimental Methods

Introduction

The numerical material models developed during this work required a number of
composite material coupon tests to generate calibration data. To validate the models a
further set of coupons were tested.

To support the calibration and validation test work, manufacturing techniques were
developed and novel tooling was designed. This technology, applicable to both
laboratory and industrial scale processes, was used for the manufacture of the
demonstrator components, described in Chapter 6.

3.2

Material Form

The material used for this study is Twintex, a commingled glass fibre reinforced
polypropylene matrix composite. Twintex is supplied in various forms including
rovings, woven fabric and preconsolidated sheets. A balanced 2 x 2 twill weave (10
ends/in. x 5 double ends/in.) preconsolidated form of the material, with 60% glass
fibres by weight, is used during the development of material models in this study.
Figure 3.1 shows preconsolidated woven fabric Twintex. The polymer in this case is
coloured black and the glass fibres appear white. In the later demonstrator component
stages of the work, described in Chapter 6, Twintex with white polypropylene was
used. The properties are unchanged between the two and the switch was based on
material availability.

Figure 3.2 shows a schematic view of the cross section of one commingled tow. Glass
fibres and polymer fibres are bundled together to facilitate rapid forming through
preheat and compression moulding.

3.3

Compression

Moulding

Commingled

Thermoplastic

Composites
A non-isothermal compression moulding process was used to manufacture all test
coupons and the demonstrator components in this work. The process involves
preheating, rapid transfer to a press tool and cooling during application of pressure,

3-1

see Figure 3.3. The processing parameters, used for compression moulding, were
taken from previous work by Wakeman [11][12].

Figure 3.4 shows a schematic representation of the microstructure of a


preconsolidated and a fully consolidated thermoplastic composite material. The small
voids present in the preconsolidated material are removed as air is expelled from the
composite during the forming process.

3.3.1 Tooling Concept


A novel flat plaque tool was developed to produce the mouldings from which all inplane test coupons were cut. The tool included a blankholder, to maintain tension in
the fibres during forming and a shear edge to minimise matrix flow from the tool
during the application of pressure and to allow co-moulding with flow materials such
as GMT. The blankholder is sprung to put a slight amount of tension in the aligned
fibre of the woven material, with the intention of giving improved and more
repeatable mechanical properties and to aid formability for more complex components
[14].

Figure 3.5 shows the lower half of the tool; the upper half consisted of a simple flat
face. The shear edge is also sprung to allow moulding of different thickness plaques
using the tool. Figure 3.6 shows the location of these springs. These features were
included on the flat plaque tool to validate the concept before application to a larger
more complex tool for the demonstrator component programme.

3.4

Material Model Calibration Data

Both the ABAQUS/Standard implementation of the model proposed by Chang et al


[79], presented in modified form in Chapter 4 and the PAM-CRASH bi-phase
composite model [81] presented in Chapter 5, require a range of material properties
for full calibration. This work combines experimental test generated data and
published values to produce the required datasets.

3-2

Table 3.1 shows the material constants and damage parameters required for the
ABAQUS composite material damage model and the source of the values used in this
work. Table 3.2 shows the material constants and damage parameters required for the
PAM-CRASH bi-phase composite material damage model and the source of the
values used.

It was important that data was accurate and gathered under standardised and
repeatable conditions. Where possible BS/ISO or ASTM standard test methods were
used. Where stress/strain behaviour is presented, the stress is the nominal stress based
on the original cross section area of the test specimen.

3.4.1 0/90 Tensile Test


The 0/90 tensile test was used to calibrate the Youngs Modulus for the fibre
direction as well as the failure strength for the ABAQUS damage model. The nominal
stress/strain curves were also used to calibrate the PAM-CRASH bi-phase model
damage parameters.

The British Standard method for determination of tensile properties of isotropic and
orthotropic fibre reinforced composite materials [114] was selected for these tests.
Nominally 4mm thick specimens were cut from plaques moulded using the method
described in section 3.3. The specimens were produced to the dimensions given in
Figure 3.7. Specimens were tested to ultimate failure in an Instron 1195 test machine
with a 50 kN load cell, shown in Figure 3.8. The specimen was mounted in shearlocking jaws and a linear extensometer, calibrated to measure up to 10% strain in the
longitudinal direction over a 50mm gauge length, was mounted across the central
section of the test specimen, see Figure 3.9. Specimens were tested at a displacement
rate of 1mm/minute.

Nominal stress/strain results from the tests are presented in Figure 3.10. Ultimate
tensile strength was averaged from the results of the eleven specimens tested and
Youngs Modulus was calculated over a 0% to 0.5% strain range and averaged for the
specimens. The mean Youngs Modulus was 12.17 GPa and the mean tensile strength
was 279 MPa. A typical failed specimen is shown in Figure 3.11.

3-3

3.4.2 0/90 Compressive Test


The ASTM standard [115] was used for testing compressive specimens. The
dimensions of the test specimens are shown in Figure 3.12. Experimental results are
taken from the work of Wan and Tham [116], who tested four specimens to failure.
The tests were performed in the Instron 1195 test machine, using the compressive test
fixture shown in Figure 3.13. Strain was measured using a single electrical resistance
strain gauge, within the 20mm gauge length. Specimens were tested at a displacement
rate of 1mm/minute.

Nominal stress/strain results from the tests are presented in Figure 3.14 and a
characteristic failure for one of the four compressive test specimens is shown in
Figure 3.15. Ultimate strength was averaged from the results of the four specimens
tested. The compressive strength was used for calibration of the ABAQUS damage
model and the full stress/strain response was used for calibration of the PAM-CRASH
model.

3.4.3 +45/-45 Tensile Test


The off-axis tensile test was performed using the same specimen geometry, test
method, fixture and machine as the 0/90 tensile test. The specimens were cut from a
plaque manufactured with the fibre directions aligned at 45 to the principal axes of
the mould tool. Nominal stress/strain results from the tests are shown in Figure 3.16.
A typical failed specimen is shown in Figure 3.17. Specimens were tested at a
displacement rate of 2mm/minute.

The results from this test were used to confirm the in-plane shear modulus from
manufacturers data. The curves were also used during the calibration of the ABAQUS
shear degradation and failure model.

3.5

Coupon Tests for Model Validation

For validation of the models described in Chapter 4 and Chapter 5, further tests were
used, where possible based on existing techniques. Four test specimen geometries
were selected. A compressive +45/-45 specimen, two tensile specimens with a stress

3-4

concentrator, 090 and +45/-45 and a compressive +45/-45 specimen with a


stress concentrator.

3.5.1 +45/-45 Compressive Test


The geometry of the test specimen for the off-axis compressive test was the same as
that used for the 0/90 compressive test. The specimens were cut from plaques
manufactured from blanks of material cut at 45 to the principal fibre direction. The
specimen was tested in the same rig as the 0/90 compressive specimen, although
strain across the gauge length was measured from crosshead displacement and not
using strain gauges. Specimens were tested at a displacement rate of 1mm/minute.

The nominal stress/strain results from the tests are shown in Figure 3.18 and a typical
failed specimen is shown in Figure 3.19. The side view of the specimen shows that, as
well as in-plane damage, out of plane buckling occurred at higher compressive strain
levels.

3.5.2 0/90 Tensile Hole in Plate Test


The geometry of the tensile hole in plate specimen is shown in Figure 3.20. The
specimen dimensions are the same as the standard tensile test, but the specimen has a
6.35mm (0.25 inch) hole drilled centrally, to act as a stress concentrator. The test
method and rig used were also the same as that used for the standard tensile tests.
Strain was measured using an extensometer across a 50mm gauge length of the
specimen. Specimens were tested at a displacement rate of 1mm/minute.

Nominal stress/strain results from the tests are shown in Figure 3.21 and a typical
failed specimen is shown in Figure 3.22. Photographs were also taken during the test
to investigate the visible progressive damage development. Figure 3.23 shows that
damage is only visible post failure, due to the rapid development of damage in this
test. Figure 3.24 shows a detailed view of the failed area of the specimen, where the
glass fibres have broken and the matrix has cracked.

3-5

3.5.3 +45/-45 Tensile Hole in Plate Test


The +45/-45 tensile hole in plate test used the same specimen geometry and test
method as the fibre direction hole in plate specimen. The strain across the 50mm
gauge length was measured using an extensometer. Specimens were tested at a
displacement rate of 2mm/minute.

Nominal stress/strain results from this test are shown in Figure 3.25 and a typical
failed specimen is shown in Figure 3.26. This specimen shows that necking occurred
around the hole during the test. This can also be seen in Figure 3.27, which shows the
specimen during the test. Clearly, strain in the specimen predominantly occurs around
the centrally placed stress concentrator. Figure 3.28 shows the failed part of the
specimen, post-test. It can be seen that little fibre failure has occurred and that the
ultimate failure mode is fibre pull-out from the polypropylene matrix.

3.5.4 +45/-45 Compressive Hole in Plate Test


A +45/-45 compressive test coupon with a stress concentrator, in the form of a
central hole was also tested. The specimen was based on the tests carried out by
Lessard and Chang [83] in their original damage model development work. The
specimen geometry is shown in Figure 3.29. The specimen was tested in a different
rig to the other compressive specimens. The rig, specifically developed by Duckett
[117], shown in Figure 3.30, end loads the specimen rather than shear loading and has
guide plates screwed on to promote in-plane damage in the specimen and minimise
buckling. The specimen was tested to a higher displacement than those in Lessard and
Changs [83] work, due to the ductile nature of the thermoplastic matrix. Specimens
were tested at a displacement rate of 1mm/minute.

Nominal stress/strain results from the tests are shown in Figure 3.31. A typical failed
specimen is shown in Figure 3.32. Like the +45/-45 compressive test specimens,
out-of-plane deformation was observed during the test. This buckling can be clearly
seen in the side view of the tested specimen.

3-6

3.6

Conclusion

Optimised non-isothermal manufacturing parameters for Twintex have been used


throughout this research to manufacture flat plaques. The flat plaques were
manufactured using a novel tool design.

Test specimen data has been obtained through a varied programme including standard
and novel test methods. This data has subsequently been used to calibrate and validate
two material models for predicting the damaging behaviour of Twintex. Fibre
direction test results for both tension and compression gave good results up to failure.
Off axis tests in compression showed some buckling, which gave uncharacteristic
stress results at large strain levels, since the buckling load for the specimens is lower
than the load required to deform the test specimens purely in-plane.

3-7

3.7

Tables

Material Constants
E11
12.17
E22
12.17
0.08

G12
1.04
G13
1.52
G23
1.52

GPa
GPa
GPa
GPa
GPa

Source
Experimental Test
Experimental Test
Manufacturers Data
Manufacturers Data
Manufacturers Data
Manufacturers Data

Damage Model Parameters


279 MPa
T1U
279 MPa
T2U
137 MPa
C1U
137 MPa
C2U
1.410-5

0.45
12U

Source
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Simulation of Experimental Test
Simulation of Experimental Test

Failed Material Constants


E11F
1 MPa
E22F
1 MPa
0 MPa
F
G12D
25 MPa
G12F
1 MPa

Source
Nominal Value
Nominal Value
Nominal Value
Simulation of Experimental Test
Nominal Value

Table 3.1

Material constants and damage parameters for ABAQUS model

3-8

Material Constants
E11T
12.17
E11C
12.17
E22T
11.40
E22C
11.40
E33T
5.3
E33C
5.3
0.08
12
0.08
13
0.36
23
G12
1.04
G13
1.52
G23
1.52

GPa
GPa
GPa
GPa
GPa
GPa

GPa
GPa
GPa

Damage Model Parameters


0.0053
iT
0.017
1T
0.039
uT
d1T
0.25
duT
0.9
0.0053
iC
0.017
1C
0.039
uC
d1C
0.50
duC
0.9

Source
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Manufacturers Data
Manufacturers Data
Manufacturers Data
Manufacturers Data
Manufacturers Data
Manufacturers Data
Manufacturers Data
Manufacturers Data
Source
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Experimental Test
Experimental Test

Table 3.2
Material constants and damage parameters for PAM-CRASH
deviatoric bi-phase model

3-9

3.8

Figures

Y (2)
Z (3)

X (1)

Figure 3.1

Preconsolidated woven commingled Twintex fabric

Z (3)
X (1)
Y (2)
Glass
Fibre

Polypropylene
Fibre

Figure 3.2

Section through commingled tow

3-10

Blank of Twintex cut to


approximate un-formed
component shape.

1.

2.

Blank heated to 220C in


infrared oven.

Rapid transfer
to pre-heated
tool at 60C.

3.
Tool closed to form
component.

4.
40 bar pressure applied
for 80 seconds.

Formed component
removed from tool and
trimmed to shape.

5.

Figure 3.3

Schematic diagram of non-isothermal compression moulding process

3-11

Preconsolidated
Twintex

Glass
Fibre

Fully Consolidated
Twintex

Glass
Fibre

Void
Polypropylene

Figure 3.4

Polypropylene

Preconsolidated and fully consolidated Twintex micro structure

Shear Edge
Blankholder

Tool Surface

Figure 3.5

Top half of combined shear edge and blankholder tooling concept

Shear Edge
Spring

Figure 3.6

Blankholder
Spring

Side of tooling concept showing shear edge and blankholder springs

3-12

50mm
25mm

250mm

Figure 3.7

Tensile tests specimen geometry

Figure 3.8

Instron 1195 test machine used for tensile coupon tests

3-13

Figure 3.9

Tensile test specimen with extensometer mounted

350
Specimen 1
Specimen 2
Specimen 3
Specimen 4
Specimen 5
Specimen 6
Specimen 7
Specimen 8
Specimen 9
Specimen 10
Specimen 11

300

Stress (MPa)

250
200
150
100
50
0
0

0.5

1.5
Strain (%)

Figure 3.10

0/90 tensile test results

3-14

2.5

Figure 3.11

Post test tensile 0/90 specimen


20mm

12.5mm

100mm
mm
Figure 3.12

Compressive tests specimen geometry

Figure 3.13

Shear loaded compressive test fixture

3-15

140
Specimen 1
Specimen 2

120

Specimen 3
Specimen 4

Stress (MPa)

100
80
60
40
20
0
0

0.2

0.4

0.6

0.8
Strain (%)

Figure 3.14

0/90 compressive test results [116]

Figure 3.15

Post test compressive 90/90 specimen [116]

1.2

1.4

80
Specimen 1

70

Specimen 2

60

Stress (MPa)

50
40
30
20
10
0
0

8
Strain (%)

Figure 3.16

+45/-45 tensile test results


3-16

10

12

14

16

Figure 3.17

Post test tensile +45/-45 specimen

45
Specimen 1

40

Specimen 2

35

Stress (MPa)

30
25
20
15
10
5
0
0

10
Strain (%)

Figure 3.18

+45/-45 compressive test results

Figure 3.19

Post test compressive +45/-45 specimen

3-17

15

20

25mm

50mm

12.5mm

125mm

6.35mm
250mm

Figure 3.20

Tensile tests specimen with stress concentrator geometry

200
Specimen 1

180

Specimen 2

160

Stress (MPa)

140
120
100
80
60
40
20
0
0.0

0.5

1.0

1.5

2.0

Strain (%)

Figure 3.21

0/90 tensile hole in plate test results

Figure 3.22

Post test tensile 0/90 specimen with stress concentrator

3-18

2.5

Figure 3.23

Damage development in 0/90 tensile hole in plate test

Figure 3.24

Typical failure in 0/90 tensile hole in plate test

3-19

120
Specimen 1
Specimen 2

100

Stress (MPa)

80

60

40

20

0
0

10

12

Strain (%)

Figure 3.25

+45/-45 tensile hole in plate test results

Figure 3.26

Post test tensile +45/-45 specimen with stress concentrator

Figure 3.27

Damage development in 0/90 tensile hole in plate test


3-20

14

Figure 3.28

Typical failure in +45/-45 tensile hole in plate test

25mm
12.5mm

6.35mm
50mm
100mm

Figure 3.29

Compressive test specimen with stress concentrator geometry

3-21

Figure 3.30

End loaded compressive test fixture

3.5
Specimen 1
Specimen 2

Specimen 3

Load (kN)

2.5
2
1.5
1
0.5
0
0

Displacement (mm)

Figure 3.31

+45/-45 compressive hole in plate test results

3-22

Figure 3.32

Post test compressive +45/-45 specimen with stress concentrator

3-23

Chapter 4
4.1

Implicit Finite Element Damage Modelling

Introduction

An implicit finite element analysis technique is usually employed for small


displacement stress analysis. An iterative solution of this type is therefore usually not
particularly suited to the large displacement analysis of composite materials, whose
post first-ply failure behaviour is often highly non-linear. The method can though be
adapted to cope with the onset of damage in composites and the consequent reduction
in material properties.

Rather than analysing a single displacement step, an incremented technique with


relatively small displacements can be used. If failure criteria are implemented in the
material model, leading to progressive degradation of elastic material properties, a
composite component can be simulated as it displaces and continuum damage
evolves. An example of this type of analysis, implementing a progressive damage
model developed by Chang and Lessard [79], is described in Chapter 2, where it is
shown that this method can been used with some success to model in-plane damage
development around a stress concentration for thermoset matrix composites.

This part of the work attempts to retain the methodology, but develop the model
proposed by Chang and apply this to thermoplastic matrix composites. Since the
Chang model was developed for unidirectional thermoset matrix composites, the
ductility of the matrix and the fibre architecture differ significantly from those used in
this study. Failure in the composite, modelled by Chang using either fibre or matrix
failure criteria, is therefore represented here using a ply based maximum strength
approach, although the shear damage methodology is retained. The modification of
the damage and failure prediction laws allows more accurate simulation of the
behaviour of a woven composite and enables the model to be calibrated with a small
range of standard coupon tests.

The model is implemented as a user defined material within the ABAQUS/Standard


implicit finite element code [118]. Failure checks and elastic property reduction are
performed by a FORTRAN subroutine subsequent to each load step.
4-1

Validation of the model is performed for a number of in-plane load cases including
tension and compression and for an in-plane specimen with a stress concentrator, both
in the fibre direction and for off axis loading.

4.2

Analysis Technique

Since the modelling strategy developed during this research was to be implemented in
larger analyses with more complex geometry than those used for laboratory scale
validation tests, a primary requirement was that the methodology used could be
efficiently scaled up to larger components. Often, techniques for modelling
progressive damage in composite materials rely heavily on highly refined meshes,
mixed shell and solid elements, multi-layered shells or complex and numerically
expensive sub models to capture behaviours such as micro-cracking and delamination
[15]. The material damage model developed here is used with standard, full and
reduced integration, shell elements. These are typical of the types of elements used in
industry for stress analysis solutions. Hence the material model can be transferred to
larger simulations without the need for costly refinement of, or alteration to, existing
meshes.

Single layer shell element models were generated using HyperMesh version 4.0.
Subsequent editing of input decks was carried out to modify the material cards and
implement the damage model.

4.2.1 Finite Element Code


The choice of an implicit finite element code for non-linear problems of this sort is
often overlooked. However through careful material model development an implicit
stepped solution has the potential to offer a computationally efficient solution when
compared to an explicit code. The ABAQUS/Standard implicit finite element code
[118] was selected for this work, as it allows a significant degree of user input within
available material models. This versatility is facilitated by the user defined material
card within the ABAQUS input deck, which gives control over the elastic material
constants used by the analysis.

4-2

4.2.2 Material Model


The material model selected for the analysis was *ELASTIC TYPE=LAMINA. The
model allows the definition of a single ply of an anisotropic lamina material using the
following elastic material constants E1, E2, 12, G12, G13, and G23.

The model includes the switch, DEPENDENCIES=3, which gives user control over
the elastic constants at the start of each load increment. 3 field variables are used to
define the values assigned to the material properties, relative to their initial,
undamaged, state. The field variables are flags, which range from 1 to 0 and define a
variation of the elastic material constants between the initial state, given in the first
line of the material model, and the minimum values described in the field variable
dependent lines of the material model.

Field variables can be set at the start of each increment by calling a USDFLD
subroutine from the *USER DEFINED FIELD card. The USDFLD routine allows
output variables from the analysis to be used to compute current values of field
variables and return these to the material model. In the case of the damage model
developed for this work, the subroutine uses stress and strain results from the analysis
to calculate accumulated damage and adjusts the field variables accordingly.

4.2.3 Element Definition and Formulation


For all the analyses performed in this section of the research, Kirchhoff thin shell
elements are used [118]. These elements were selected as they can be used to
accurately model the behaviour of structures, which have a section thickness that is
small relative to the overall dimensions of the component. Test specimens modelled
are typically over 100mm long and have a section thickness of less then 4mm. The
final demonstrator component, described in chapter seven, is over 1000mm in length
and has a maximum section thickness of 8mm. In practice, considering the limitations
of the non-isothermal compression moulding technique used in this research, it would
be unlikely that a component could be produced for which a shell element model was
not the most suitable analysis technique.

4-3

The shell elements used, are, in the case of quadrilateral elements, defined by four
nodal points. These nodes describe the spatial position of the element and also, using
the right-hand rule, define the normal direction of the shell. This normal is critical
for correct definition of laminate ply sequences. During the calibration and validation
of the model a comparison of the full and reduced integration elements available in
ABAQUS was made. All test specimen simulations were run with both element types,
those being S4 and S4R. Where the element code relates to the following:

S4 - Stress/displacement element, 4 nodes

and:

S4R - Stress/displacement element, 4 nodes, Reduced integration.

ABAQUS stress/displacement shell elements use a Lagrangian formulation where the


element displaces according to the behaviour of the constituent material. Guassian
Quadrature is used to solve for static equilibrium at each integration point within the
element.

4.2.4 Section Definition


Shell elements are defined in ABAQUS using the *SHELL SECTION card. This card
specifies the shell element thickness, material and number of through thickness
integration points. The optional COMPOSITE parameter allows the user to define
discrete layers and orientations. Using this type of section definition a complex,
multiple layered, laminate can be accurately described and assigned to shell elements.

Details of the various ply and orientation section definitions, used in the analysis of
each of the in-plane damaging specimens simulated, are given later in this chapter.

4.2.5 Definition of Orthotropy Direction


The orientation of the fibre direction for each composite material ply can be defined
independently. The *ORIENTATION card is used to, with either a local or the global
coordinate system, assign the fibre directions of the woven composite. In this case, for
the analysis of flat rectangular test specimens, all materials are oriented relative to the
4-4

global coordinate system. If the model were to be used for a spherical or cylindrical
component a local coordinate system could be defined and the composite oriented
accordingly. This global coordinate orientation technique, when used in conjunction
with shell elements, identifies the orientation direction which lies closest to the plane
of each shell and projects the principle fibre directions to that plane accordingly. This
feature is used extensively in the analysis performed on demonstrator components,
detailed in Chapter 6.

4.3

Development of the Damage Model

The model was developed to represent the in-plane damaging behaviours of a


balanced weave composite material. Three critical, ply level, damage modes were
identified through review of previous work. These were tensile, compressive and
shear damage. These modes represent, at a ply level, more complex micro damage
accumulation. For example, in the case of tensile damage, degradation of material
properties could be the result of matrix cracking, fibre matrix debonding, fibre failure
or a combination of all three. The model, although not identifying these discrete
damage types, represents the accumulated effect of them through calibration from
physical test. The assumption is therefore made that the type of damage seen in
laboratory tests is similar to that developed in a more complex geometry under large
displacement loading. The initial testing and simulation and subsequent validation
work substantiates this hypothesis.

The model implements numerical representations of the global ply damage and failure
modes identified. Within the user defined damage model subroutine, these criteria are
used to directly alter the elastic material properties within the composite material
model.

4.3.1 Tensile Failure Model


In this model, the basic anisotropic linear elastic tensile behaviour of the standard
ABAQUS composite material is modified, by including a failure check. It was felt
after analysis of results from tensile test specimens that degradation of the modulus
prior to failure was not required, since the behaviour of Twintex is dominated by the
glass fibres. This leads to a characteristic linear elastic behaviour with catastrophic
4-5

failure caused by fibre breakage. To simulate this, a maximum stress criteria is


implemented in the fibre direction, which when exceeded results in a reduction of the
Youngs modulus in both the 1 and 2 direction to a nominal value of 1MPa and the
reduction of Poissons Ratio to 0. This represents the loss of local tensile load
carrying capacity in both directions when fibre failure occurs in a composite ply (eqn.
3.1).

Failure identified if:

11T > T 1U

(3.1)

Since Twintex is a balanced weave composite, the behaviour in the two principal inplane directions is identical. Hence the tensile failure check in the 2 direction also
uses a maximum stress criteria (eqn. 3.2), which in this case reduces the elastic
constants in the same way as if tensile failure in the 1 direction were identified.

Failure identified if:

22T > T 2U

(3.2)

4.3.2 Compressive Failure Model


The Youngs modulus of the composite ply applies to both the tensile and
compressive elastic behaviour of the material. Therefore, if tensile failure has
occurred and the corresponding modulus has been reduced to a nominal value, the
composites ability to carry compressive load is also compromised. In turn if
compressive failure has been identified the compressive and tensile load carrying
capacity is reduced, again to a nominal modulus of 1MPa. The compressive failure of
Twintex if evaluated in-plane, i.e. buckling effects are discounted, can also be
represented as a linear elastic behaviour followed by catastrophic failure. Hence the
compressive failure model uses, like the tensile model, a maximum stress criteria.
This criteria is used in both the 1 and 2 directions of the composite ply (see eqn. 3.3
and eqn. 3.4).

Failure identified if:

11C > C1U

(3.3)

Failure identified if

22C > C2U

(3.4)

4-6

4.3.3 Shear Damage and Failure Model


The shear damaging behaviour of the composite is more complex than the relatively
simple linear elastic followed by ultimate failure of the tensile and compressive
models. When the woven ply is loaded in shear, matrix damage is accumulated over a
large strain range before ultimate failure. It is therefore not possible to use a
maximum stress based failure criteria to describe the ply.

Following Changs approach [79], the shear properties of the material are degraded as
a cubic function of the shear stress. For each load/displacement increment, at each
integration point, the value of the shear stress is used to adjust the in-plane shear
modulus, G12 (see eqn. 3.5). In the model, is a fitting parameter and defines the nonlinear relationship between shear stress and strain.

12 =

12
G12

+ 123

(3.5)

In the model, G12 tends to a value of 25MPa, denoted G12D. This modulus when
reached is used up to the point at which ultimate ply shear failure is identified. The
value of this damaged, but not failed, shear modulus was arrived at through
calibration from experimental test.

In Changs work an ultimate shear failure criteria is included, based on the fibre
buckling strength and laminate shear strength. In this study, where calibration from a
basic suite of experimental tests was a primary requirement, a maximum shear strain
criteria is used (see eqn. 3.6). When this failure is identified, shear loads can no longer
be carried, therefore the in plane shear modulus, G12, is reduced to a nominal value of
1MPa.

Failure identified if:

12 > 12U

(3.6)

4.3.4 Implementation
As mentioned previously, the material model includes the DEPENDENCIES=3
switch, which allows the elastic material properties to varied using user defined field
variables. Within the model these dependency variables are returned at each step of
the analysis by calling the USDFLD subroutine. This routine is written using the
4-7

FORTRAN programming language and implements the mathematical damage and


failure checks of the model. Appendix A gives the code used for the model and
includes the tensile compressive and shear failure checks and the shear damaging
behaviour. The form of the mathematical implementation of the shear damage model
is arrived at through manipulation of Equation 3.5. A derivation of the shear damage
formulation is given in Appendix B.

The field variables used for the variation of the material properties are output during
the analysis and identify damage or failure in the specimen. Field variable 1 (FV1)
denotes tensile or compressive fibre direction failure. This variable is set initially at 0,
no failure, and rises to 1 when ultimate failure is identified. Similarly, field variable 2
(FV2), is initially 0 and becomes 1 when shear failure is identified. Field variable 3
(FV3) relates to shear damage and varies between 0 and 1 as shear damage
accumulates during the analysis. Figure 4.1 summarises the implementation of the
damage model as a FORTRAN subroutine in the ABAQUS/Standard finite element
package.

4.4

Calibration

To calibrate the damage model, material properties were taken from Saint-Gobain
Vetrotex published information [18] and calibration tests. The tests were used to
confirm manufacturers information for the processing techniques employed and to
identify the material ultimate strengths and fitting parameters required for the model.
A detailed description of experimental test work and a summary of the material
properties, failure criteria and calibration parameters used in the model are given in
Chapter 3. The calibrated input cards for the ABAQUS field dependant material are
given in Appendix C. Throughout this section of the work, simulation results, like the
experimental results in Chapter 3, are presented as nominal stress/strain.

4.4.1 Simulation of Calibration Coupon Tests


The tests used to derive the model parameters were simulated using the damage
model to confirm that calibration coefficients and material properties were correct and
that the experimentally identified failure mode was being predicted in the correct
location and at the correct nominal stress and strain level.
4-8

4.4.2 Tensile 0/90 Test Simulation


The finite element mesh used to model the 0/90 tensile coupon tests is shown in
Figure 4.2. 170 reduced integration four node shell elements were used, with a
nominal edge length of 5mm. The aspect ratio of the elements in the areas where the
specimen is gripped in the test is increased, since these nodes were either constrained
in all degrees of freedom at the fixed end, or constrained with a constant displacement
boundary condition at the moving end. This boundary condition arrangement was felt
to be the most accurate representation of the shear loading jaws used to clamp the
coupon during experimental test.

The simulation was set to run to a total displacement of 5mm, which equates to a
strain level of approximately 3.33% in the gauge length of the specimen. The
increment for each analysis step was set to 0.025mm or 0.0167% strain. This
increment was selected to allow the model to accurately identify the point at which
damage starts to occur in the test. Larger increment sizes could lead to failure being
identified at an incorrect displacement due to the resolution of the solution. Further
investigation of the sensitivity of the model to increment size was undertaken and a
minimum solution resolution is recommended in section 4.6.1.

Element stress/strain and field variable status output from the analysis was requested
after every fifth increment. This was chosen to minimise the size of output files from
the analysis, since reaching the disk quota limit before completion of an analysis
would cause premature termination. Nodal output was requested every second
increment, resulting in 100 data-points for a full tensile test simulation run. This was
felt to be sufficient to provide output for comparison with experimentally derived
stress/strain curves.

Since the experimentally derived stress/strain behaviour in the fibre direction is


predominantly linear elastic with a catastrophic failure, the finite element simulation
curve shows good correlation, see Figure 4.3. The strain at failure for the simulation is
slightly lower than experiment, since in the test, material behaviour is not truly linear
elastic throughout. As the coupon nears the ultimate stress of the material, there is a
slight reduction in modulus. This is observed from approximately 1% strain where the
experimental curve dips below the simulation result. This effect is negligible though
4-9

when considering the spread of failure strain results from experiment. The calibration
for tensile failure is therefore shown to be accurate and the model successfully
predicts the behaviour of Twintex in tension.

The final damaged state of the specimen is shown for test and simulation to compare
the damage prediction, see Figure 4.4. In this case it is not possible to visually draw
conclusions from the result as the final plot state of the field variable for tensile
damage is prior to the failure event. Once failure has occurred a large amount of
element deformation takes place and a static equilibrium solution cannot be
converged and hence a plot of the failed element(s) cannot be generated.

4.4.3 Compressive 0/90 Test Simulation


The finite element mesh used to model the compressive coupon tests is shown in
Figure 4.5. 224 reduced integration four node shell elements were used, with a
nominal edge length of approximately 1.5mm within the gauge length area of the
specimen. Although the full specimen is modelled, the majority of the specimen in
this case is constrained, since the loading is applied with locking jaws which transmit
the compressive load to the specimen. The gauge length of the specimen is minimised
to alleviate the problem of buckling as the specimen is loaded.

The compression test results in Figure 4.6 show good correlation between FE and
experimental stress/strain behaviour. The form of the curve is similar to that seen in
the tensile test. The behaviour is predominantly linear elastic with an ultimate failure.
Figure 4.7 shows the predicted damage at the end of the simulation. Compressive
damage is predicted throughout the gauge length of the specimen. The damage seen in
test shows a line of failure in the gauge length of the specimen.

4.4.4 Tensile +45/-45 Test Simulation


+45/-45 tensile tests were used to calibrate the shear degradation part of the damage
model. Figure 4.2 shows the mesh used for simulation of these specimens. This mesh
is identical to that used in the 0/90 tensile test simulations, although the material is
oriented differently.

4-10

Figure 4.8 shows the comparison between experimentally observed stress/strain in the
gauge length and simulation result. The fitting parameter, , and the damaged shear
modulus was calibrated iteratively using this simulation and test result. These
specimens exhibited a much larger strain to failure than the 0/90 tensile specimens,
due to gradual reduction of the shear modulus during the test. The results show that at
strains of up to 20% in the gauge length, the implicit elements and the shear
degradation model can reproduce the shear properties of the woven material. Figure
4.9 shows a comparison between an experimentally damaged specimen and the
predicted damage from simulation. The simulation plot shows that the specimen
damages is shear from the early stages of the simulation, but does not fail in a
catastrophic manner until very high strain levels. During this calibration process, the
in plane shear modulus, G12, was also adjusted from 1.20GPa to 1.04GPa to improve
the correlation between simulation and test.

The ultimate shear failure, which is also calibrated from this experimental test, shows
that the simulation predicts failure at approximately 19% strain in the gauge length of
the specimen.

4.5

Validation

Various coupon tests were simulated to validate the damage model calibration. These
simulations are classed as validations, since no alteration of properties or damage
parameters was undertaken after the calibration from the three simple tests detailed in
sections 4.4.2 to 4.4.4. These specimens were similar to those used for the calibration,
but included a stress concentration in the form of a circular hole, except in the case of
the +45/-45 compressive specimen, which did not include a stress concentration.

4.5.1 Compressive +45/-45 Test Simulation


The compressive +45/-45 simulation uses the same finite element mesh as the
0/90 compressive specimen, see Figure 4.5. Since the parameters describing the
shear behaviour of the material were calibrated from the off axis tensile test, there
were no parameters taken from this test specimen.

4-11

The stress/strain response from experiment and simulation are shown in Figure 4.10.
The correlation between the curves is not as good as that seen in previous tests, and
the simulation is predicting the specimen to be stiffer than is actually observed in
experiment. This phenomenon can be explained by observing the damaged specimen
and predicted damage in the model, see Figure 4.11. The simulation shows a steady
build up of in-plane shear damage, whereas the test specimen has clearly deformed
out of the plane of the test. This out of plane deformation explains why the
experimentally derived curve shows a lower stress level than the simulation result.

The simulation result also shows a stepped response. This is caused by fibre
compressive failure occurring in the model and resultant reduction of the local fibre
direction modulus of the specimen in certain elements. The location of this failure can
be seen in Figure 4.30, where a similar behaviour is observed in the full integration
shell element model of the test. This stepped behaviour continues to occur after 5%
strain in the gauge length of the specimen.

This result shows that the model is accurate for in-plane damage modes, but when
material deformation occurs out of plane, the model cannot accurately capture the
behaviour. This shows that the model should be used with care if out of plane,
buckling deformation is expected.

4.5.2 Tensile 0/90 Hole in Plate Test Simulation


Figure 4.12 shows the finite element mesh used to simulate the tensile hole in plate
specimen. The specimen was manufactured exactly as the plain tensile test specimens,
but subsequently a hole was drilled to represent initial damage.

Figure 4.13 shows the stress/strain response over the central 500mm gauge length for
the test and simulation result. The simulation shows good correlation with
experiment, although damage and subsequent ultimate failure is predicted to occur at
a lower stress level than that seen in experiment. This is due to a whole element being
failed when stress levels around the hole reach the failure strength of Twintex. In the
experimental test, the stress concentrator causes only a small amount of failure,
initially at the edge of the hole, which does not fail the specimen catastrophically.
This can be seen in the experimental curve as a slight reduction in modulus and
4-12

shallowing of the stress/strain curve. In the simulation, as soon as failure has occurred
in one element, load is transferred and the surrounding elements also fail
catastrophically.

The damage variable plot and failed test specimen, shown in Figure 4.14, highlight
this catastrophic failure seen in the simulation, where at the final step of the analysis a
whole group of elements has failed during the same iteration. This large failure occurs
at a higher load in the experimental test, but as expected, in the same location, around
the stress concentration.

4.5.3 Tensile +45/-45 Hole in Plate Test Simulation


The +45/-45 tensile test with a stress concentrator uses the same mesh as the fibre
direction test, see Figure 4.12. As in the fibre direction test, the shear damage part of
the model over predicts damage around the hole for the tensile stress concentrator
situation. This is shown by Figure 4.15 where the experimental stress/strain plot
shows a much stiffer response than the simulation. In this case, like the previous
0/90 specimen with a stress concentrator, the model over predicts damage, resulting
in reduced stiffness in the simulation. This result could also be explained by the
inability of the model to account for the reorientation of fibres in areas of high shear
strain. This could potentially lead to a stiffening effect around the hole, which is not
modelled accurately. This phenomenon is discussed in more detail in Chapter 7.

The damage contours from the model, in Figure 4.16, show shear damage developing
throughout the specimen and particularly around the hole from the early stages of the
simulation. The damage model predicts a distribution of damage throughout the
specimen, with necking occurring along the entire length. This was not seen in the
experimental test, where the necking phenomenon was only observed around the area
of the hole.

4.5.4 Compressive +45/-45 Hole in Plate Test Simulation


The mesh used for the +45/-45 compressive test coupon with a stress concentrator is
shown in Figure 4.17. Since the test used an end loading rig, the specimen is modelled
with constraints only along the edge nodes at one end, with displacement applied at
the other. Tests results are given as load/displacement, since in this test no strain
4-13

gauge or extensometer was used, see Figure 4.18. The correlation between test and
experiment is good up to 2mm displacement, where the experimental test shows a
reduction in load, that is not reflected in the simulation. This is explained by the
specimen deformation seen in experiment, see Figure 4.19. During test, the Twintex
deforms out of plane in a buckling mode, which is not accounted for by the damage
model. The load therefore begins to reduce after 3.5mm displacement, whilst the
simulation result suggests that the damage will gradually accumulate as the load
increases.

The lack of correct treatment of fibre reorientation in areas of high shear strain,
mentioned in section 4.5.3, when discussing the under-stiff response of the model for
the tensile test, could also be used to explain the over-stiff response observed for this
simulation. Again, this is discussed in more detail in Chapter 7.

4.6

Model Sensitivity

The model was developed and calibrated from the results of experimental test. The
sensitivity of the model to parameters such as Youngs Modulus and tensile and
compressive failure strain taken from these tests is predictable. For example
increasing the tensile failure stress in the damage model will result in the model
failing at a higher stress level. Similarly, variation of the material constants results in
a predictable change in the linear elastic behaviour of the model. There are however
parameters in the model for which the sensitivity is not as easy to estimate. The
variables that were investigated can be split into two distinct groups, material model
constants and analysis parameters.

The constants that were investigated, were both part of the shear damage model, since
the fibre direction damage models are calibrated with exact values taken directly from
test results. These parameters were, , the shear damage parameter and the damaged
shear modulus G12D.
The analysis parameters that were investigated were, the number of analysis steps, the
element formulation (full or reduced integration) and mesh refinement.

4-14

4.6.1 Sensitivity to Analysis Step


The sensitivity to analysis step was investigated for the +45/-45 tensile test, since
this test undergoes a large amount of deformation, with progressive damage
throughout. The analysis was set to run to a total displacement of 30mm, which
equates to a total strain of approximately 20%. Five analyses were run with a total
number of steps of 20, 50, 100, 200 and 500, or approximate steps of 1%, 0.4%, 0.2%,
0.1% and 0.04% strain. Stress/strain responses from each of these simulations were
plotted and compared to experimental test data, see Figure 4.20. The results are shown
up to 10% strain to highlight the region of the analysis where the result is most
sensitive to step size.

It can bee seen from the results that the solution has converged by the 200 step
simulation although the 100 step solution shows a satisfactory level of convergence.
The step size of 0.1% strain, from the 200 step solution, was selected as a minimum
for the analysis of Twintex.

4.6.2 Sensitivity to Shear Non-Linearity Parameter


The +45/-45 tensile test simulation was also used to investigate the sensitivity of the
model to , the shear damage parameter. This sensitivity analysis was used as part of
the calibration of the model, since the result was used to select the value of used for
all simulations.

Figure 4.21 shows the result of the five sensitivity simulations run against the result
from experimental test. The five values used were 1.410-6, 6.310-6, 1.410-5,
6.310-5 and 1.410-4. The stress/strain curves show that the region between 0.5% and
2% strain is the area that is most sensitive to the shear damage parameter calibration.
As the magnitude of the parameter is reduced, the magnitude of the identified damage
and hence the reduction of shear modulus is also reduced. From this result, an value
of 1.410-5 was selected as most accurately representing the relationship between
damage and modulus reduction for Twintex undergoing shear deformation.

4-15

4.6.3 Sensitivity to Calibration of Damaged Shear Modulus


The third of the calibration parameters investigated was G12D, the damaged shear
modulus. Again, five values were selected to investigate the sensitivity in the +45/45 tensile test simulation. The values chosen were 5MPa, 15MPa, 25MPa, 45MPa
and 60MPa.

Figure 4.22 shows the stress/strain results from this sensitivity analysis, compared to
the experimental test result, up to a strain level of 10%. The plot shows that the region
of the analysis that is sensitive to the damaged shear modulus, G12D, is above
approximately 2% strain. The best correlation between experiment and analysis is for
a G12D value of 25MPa. This value was selected and used for all simulations.

4.6.4 Sensitivity to Element Formulation


An in depth investigation into the effect of element formulation was also undertaken.
All calibration and validation simulations, originally run with reduced integration
shell elements were simulated with full integration elements. The meshes used were
identical to the original simulations.

Figure 4.23 shows the stress/strain result for the 0/90 tensile test. The simulation
shows almost identical response, although the full integration elements predict failure
slightly before the reduced integration and the analysis remains stable after first
failure is predicted. This is shown graphically by the damage contour plot in Figure
4.24, where the full integration simulation gives damage output showing the areas
where failure has been identified. The elements showing failure levels of 0.25 and 0.5
are average values for the four integration points of the element. A smoothed contour
plot was not used, since this would not allow easy identification of fully failed
elements.

The results from the compressive 0/90 simulation show a similar result to the tensile
test. Figure 4.25 shows that the stress/strain response is identical for both the full and
reduced integration elements. The damage plot, Figure 4.26, shows a slight difference
in the level of damage identified and the shape of the damage area, when comparing
the two simulations. The full integration elements predict damage slightly earlier than
the reduced integration, in the elements adjacent to the area where the boundary
4-16

conditions are applied. This is due to the stress concentration effect of the nodal
constraints used.

The effect of element integration is more marked for the simulations where shear
damage is being predicted. This is shown by Figure 4.27. The +45/-45 tensile test
simulation predicts ultimate failure sooner with full integration, at around 16% strain,
compared to 19% strain in the reduced integration simulation. The plot comparing
field variable 3 for the two simulations, Figure 4.28, shows that the prediction of
damage development in the early stages of the simulation is very similar for the two
types of element formulation.

The result for the +45/-45 compressive test simulation shows the most significant
difference between the two formulations, see Figure 4.29. The stress/strain response is
similar up to approximately 2% strain, where the responses diverge considerably. In
the full integration simulation, much more severe damage is predicted. This damage is
shown in Figure 4.30. The full integration elements, as well as identifying a similar
level of shear damage, seen in the comparison of field variable 3, show compressive
fibre direction damage highlighted by the plot of field variable 1.

The tensile 0/90 specimen with a stress concentrating hole also gives a differing
result dependant on the type of element used. The stress/strain plot, Figure 4.31,
shows that in the case of full integration, damage is predicted at just over 1% strain,
whereas in the reduced integration elements this is not predicted until 1.5% strain.
The damage in the full integration elements occurs earlier due to the proximity of the
integration point to the stress concentration. Higher stress means that failure is
reached sooner in the analysis. Figure 4.32 shows that the full integration elements
remain stable and converge a solution after damage has occurred. The reduced
integration solution fails to converge soon after the first point at which damage is
predicted to occur.

The +45/-45 tensile specimen with a stress concentration models under-predict the
stiffness, with either element type, see Figure 4.33. Figure 4.34 shows the field
variables representing damage during the simulation. The full integration solution

4-17

predicts damage and failure sooner than the reduced integration elements. Again this
is due to the proximity of the integration points to the stress concentration.

Figure 4.35 shows the stress/strain response from the +45/-45 compressive
specimen with a hole stress concentration. The results are similar to previous
observations, where the full integration elements predict a slightly higher level of
damage. At around 4mm displacement the response of the full integration simulation
shows a drop in load, which is not seen in the reduced integration simulation. The
comparative plot of field variable 3 for each simulation highlights this difference in
damage prediction, see Figure 4.36. In both cases the solutions do not predict the drop
off in load seen as the specimen starts to deform out of plane. As explained earlier,
during the validation section of the work, this phenomenon is not simulated by the
model.

4.6.5 Sensitivity to Mesh Refinement


The sensitivity of the solution to mesh refinement was also investigated. Figures 4.37
and 4.38 show two meshes used to simulate the +45/-45 compressive specimen with
a stress concentration. This coarse and fine mesh were compared to the standard
density mesh used during the validation of the model. Figure 4.39 shows the
stress/strain response for the coarse mesh. Both full and reduced integration solutions
are shown, since the mesh density also affects sensitivity to element formulation. In
both cases the simulations predict shear damage, although it is noted that the full
integration solution does not show the load drop off at 4mm predicted by the medium
density mesh. The full integration elements also show a significantly higher stiffness
than their reduced integration counterparts. The field variable 3 plots in Figure 4.40
also show a significant difference in the level and pattern of damage.

The fine mesh results in Figure 4.41 show that when the mesh is refined, the effect of
the number of integration points on the stiffness of the elements is far lower. The
load/displacement response up to 3.5mm is very similar for both models. The full
integration solution does though predict a load drop off due to damage development
in the specimen, at approximately 3.5mm, which is not seen in the reduced integration
solution. A comparison of field variable 3 during the simulation is shown in Figure
4.42. The level of shear damage predicted by the full integration solution is slightly
4-18

higher, although the difference is not as marked as that displayed by the coarser
meshes.

In general for a reduced integration element formulation, the load/displacement


response does not appear to be particularly sensitive to element formulation. Even the
coarse mesh gives a good representation of the specimens behaviour up to the point at
which out of plane damage begins to occur. The main difference between the meshes
can be seen when comparing the field variable plots, especially for the full integration
solution.

4.7

Conclusion

A damage model based on a maximum stress fibre direction failure criteria, a strain
based shear failure and a shear degradation behaviour has been developed and
calibrated for woven Twintex. The calibration method suggested can be performed
and the material fully characterised with only 3 test geometries.

Modelling of calibration and validation specimens has shown that the model is
accurate when predicting the behaviour of Twintex for situations where loading is in
the principal fibre directions. The shear behaviour of the material can also be
accurately simulated, although when shear deformation is unevenly distributed in a
specimen such as the tensile +45/-45 test with a stress concentration, the model
under-predicts the stiffness. The only test specimen behaviour that the model cannot
accurately predict is out of plane buckling. This is seen in some of the compressive
tests, where the in-plane model over-predicts the stiffness of test specimens.

A further limitation of the model for small coupon tests, such as those presented in
this chapter, is the ability of an implicit code to converge an equilibrium solution after
material properties have been significantly reduced. The property reduction results in
high levels of element deformation and subsequent failure of the analysis, not just the
material model.

4-19

The model has been shown to be sensitive to calibration parameters, solution step
size, element formulation and mesh density. The calibration parameters and step size
were found to be the most critical of these variables.

The results from this piece of work demonstrate that in general, prediction of in-plane
damage development is possible for a woven composite material, using a stepped
implicit finite element solution, with a basic damage model.

4-20

4.8

Figures

First Increment
of Load

Completed Analysis
Yes
NEXT Increment
of Load

No

Total Load
Reached?

Calculated
Displacements,
Stresses

Shear Properties
Degraded

Element Material
Properties Degraded

Element Material
Properties Unchanged

Yes
No
Failure Criteria Checks
Stress/Strain at Integration Points

FAILURE

Figure 4.1
Schematic of damage model implementation as an ABAQUS user
defined FORTRAN subroutine

4-21

Figure 4.2

Tensile test - shell element mesh

Figure 4.5

Compressive test - shell element mesh

Figure 4.12

Tensile hole in plate test - shell element mesh

Figure 4.17

Compressive hole in plate test - shell element mesh

Figure 4.37

Compressive hole in plate test - shell element mesh (coarse)

Figure 4.38

Compressive hole in plate test - shell element mesh (refined)

4-22

300

Experimental Test
ABAQUS Model

250

Stress (MPa)

200

150

100

50

0
0

0.5

1.5

2.5

Strain (%)

Figure 4.3

0/90 tensile test simulation - stress/strain

FV1 - 2.3%

Figure 4.4

0/90 tensile test simulation - experimental and predicted damage

4-23

160

Experimental Test
140

ABAQUS Model
120

Stress (MPa)

100
80
60
40
20
0
0.0

0.2

0.4

0.6

0.8

1.0

1.2

Strain (%)

Figure 4.6

0/90 compressive test simulation - stress/strain

FV1 - 1.125%, 1.150%

Figure 4.7
damage

0/90 compressive test simulation - experimental and predicted

4-24

80

Experimental Test
70

ABAQUS Model
60

Stress (MPa)

50
40
30
20
10
0
0

10

12

14

16

18

Strain (%)

Figure 4.8

+45/-45 tensile test simulation - stress/strain

FV3 - 4%, 15%

Figure 4.9

+45/-45 tensile test simulation - experimental and predicted damage

4-25

20

80

Experimental Test
70

ABAQUS Model
60

Stress (MPa)

50
40
30
20
10
0
0

10

12

14

16

18

20

Strain (%)

Figure 4.10

+45/-45 compressive test simulation - stress/strain

FV3 - 1%, 2%, 3%, 4%, 5%, 6%

Figure 4.11
damage

+45/-45 compressive test simulation - experimental and predicted

4-26

200
180

Experimental Test

160

ABAQUS Model

Stress (MPa)

140
120
100
80
60
40
20
0
0.0

0.5

1.0

1.5

2.0

2.5

Strain (%)

Figure 4.13

0/90 tensile hole in plate test simulation - stress/strain

FV1 - 1.45%, 2.2%

Figure 4.14
damage

0/90 tensile hole in plate test simulation - experimental and predicted

4-27

100

Experimental Test

90

ABAQUS Model

80

Stress (MPa)

70
60
50
40
30
20
10
0
0

10

12

14

16

Strain (%)

Figure 4.15

+45/-45 tensile hole in plate test simulation - stress/strain

FV3 - 2.5%, 5%, 7.5%, 10%, 12%

Figure 4.16 +45/-45 tensile hole in plate test simulation - experimental and
predicted damage

4-28

5
Experimental Test
ABAQUS Model

Load (kN)

0
0

Displacement (mm)

Figure 4.18 +45/-45


load/displacement

compressive

hole

in

plate

test

simulation

FV3 - 1mm, 2mm, 3mm, 4mm, 5mm

Figure 4.19 +45/-45 compressive hole in plate test simulation - experimental and
predicted damage
4-29

60

Stress (MPa)

50

40
30

Experimental Test
20 Steps

20

50 Steps
100 Steps

10

200 Steps
500 Steps

0
0

10

10

Strain (%)

Figure 4.20

+45/-45 tensile test simulation - sensitivity to analysis steps

110

Experimental Test

100

Alpha = 1.4D-6

Stress (MPa)

90

Alpha = 6.3D-6

80

Alpha = 1.4D-5

70

Alpha = 6.3D-5
Alpha = 1.4D-4

60
50
40
30
20
10
0
0

Strain (%)

Figure 4.21 +45/-45 tensile test simulation - sensitivity to shear damage


parameter alpha -

4-30

80

Experimental Test
70

G12D = 5
G12D = 15

Stress (MPa)

60

G12D = 25
G12D = 45

50

G12D = 65
40
30
20
10
0
0

10

Strain (%)

Figure 4.22 +45/-45 tensile test simulation - sensitivity to damaged shear


modulus - G12D

4-31

300

Experimental Test
Reduced Integration

250

Full Integration

Stress (MPa)

200

150

100

50

0
0

0.5

1.5

2.5

Strain (%)

Figure 4.23

0/90 tensile test simulation - sensitivity to element formulation

FV1 - 2.2%

Reduced Integration

FV1 - 2.5%

Full Integration

Figure 4.24

0/90 tensile test simulation - predicted damage

4-32

160

Experimental Test
140

Reduced Integration
Full Integration

Stress (MPa)

120
100
80
60
40
20
0
0.0

0.2

0.4

0.6

0.8

1.0

1.2

Strain (%)

Figure 4.25

0/90 compressive test simulation - sensitivity to element formulation

FV1 - 1.125%, 1.15%

Reduced Integration

FV1 - 0.75%, 1.125%, 1.5%

Full Integration

Figure 4.26

0/90 compressive test simulation - predicted damage

4-33

80

Experimental Test
Reduced Integration
Full Integration

70
60

Stress (MPa)

50
40
30
20
10
0
0

10

12

14

16

18

Strain (%)

Figure 4.27

+45/-45 tensile test simulation - sensitivity to element formulation

FV3 - 4%, 15%

Reduced Integration

FV3 - 4%, 15%

Full Integration

Figure 4.28

+45/-45 tensile test simulation - predicted damage

4-34

20

80
Experimental Test
70

Reduced Integration
Full Integration

Stress (MPa)

60
50
40
30
20
10
0
0

10

12

14

16

18

20

Strain (%)

Figure 4.29
formulation

Figure 4.30

+45/-45 compressive test simulation - sensitivity to element

FV1 - 1%, 2%, 3%, 4%, 5%

FV3 - 1%, 2%, 3%, 4%, 5%

Reduced Integration

Reduced Integration

FV1 - 1%, 2%, 3%, 4%, 5%

FV3 - 1%, 2%, 3%, 4%, 5%

Full Integration

Full Integration

+45/-45 compressive test simulation - predicted damage


4-35

200
Experimental Test

180

Reduced Integration

160

Full Integration
Stress (MPa)

140
120
100
80
60
40
20
0
0.0

0.5

1.0

1.5

2.0

2.5

Strain (%)

Figure 4.31
formulation

0/90 tensile hole in plate test simulation - sensitivity to element

FV1 - 1.45%, 2.2%

FV3 - 0.5%, 1%, 1.45%, 2.2%

Reduced Integration

Reduced Integration

FV1 - 1%, 1.1%, 2.15%

FV3 - 0.5%, 1%, 2.15%

Full Integration
Full Integration

Figure 4.32

0/90 tensile hole in plate test simulation - predicted damage

4-36

100
90

Experimental Test
Reduced Integration

80

Full Integration

Stress (MPa)

70
60
50
40
30
20
10
0
0

10

15

20

Strain (%)

Figure 4.33
formulation

+45/-45 tensile hole in plate test simulation - sensitivity to element

FV1 - 2.5%, 12%

FV3 - 2.5%, 5%, 7.5%, 10%, 13%

Reduced Integration

Reduced Integration

FV1 - 2.5%, 12%

FV3 - 2.5%, 5%, 7.5%, 10%, 13%

Full Integration

Full Integration

Figure 4.34

+45/-45 tensile hole in plate test simulation - predicted damage

4-37

5
Experimental Test
Reduced Integration

Load (kN)

Full Integration

0
0

Displacement (mm)

Figure 4.35 +45/-45 compressive hole in plate test simulation - medium mesh sensitivity to element formulation

FV3 - 1mm, 2mm, 3mm, 4mm,


5mm

FV3 - 1mm, 2mm, 3mm, 4mm,


5mm

Reduced Integration

Full Integration

Figure 4.36 +45/-45 compressive hole in plate test simulation - medium mesh predicted damage

4-38

5
Experimental Test
Reduced Integration

Load (kN)

Full Integration

0
0

Displacement (mm)

Figure 4.39 +45/-45 compressive hole in plate test simulation - coarse mesh sensitivity to element formulation

FV3 - 1mm, 2mm, 3mm, 4mm,


5mm

FV3 - 1mm, 2mm, 3mm, 4mm,


5mm

Reduced Integration

Full Integration

Figure 4.40 +45/-45 compressive hole in plate test simulation - coarse mesh predicted damage

4-39

5
Experimental Test
Reduced Integration

Load (kN)

Full Integration

0
0

Displacement (mm)

Figure 4.41 +45/-45 compressive hole in plate test simulation - fine mesh sensitivity to element formulation

FV3 - 1mm, 2mm, 3mm, 4mm,


5mm

FV3 - 1mm, 2mm, 3mm, 4mm,


5mm

Reduced Integration

Full Integration

Figure 4.42 +45/-45 compressive hole in plate test simulation - fine mesh predicted damage

4-40

Chapter 5
5.1

Explicit Finite Element Damage Modelling

Introduction

Explicit finite element codes are used to solve non-linear or dynamic analysis
problems. This type of code is therefore ideal for the analysis of composite
components where material behaviour can be highly non-linear and large
deformations occur.

Commercial codes often include material models specifically developed for the
analysis of composites, which include damage models that the user can calibrate. In
this part of the work, experimental test and manufacturers published data is used to
calibrate one such material model. The model is then validated with the same test
specimen data used in the previous, implicit finite element analysis chapter. These
include fibre direction and off axis test coupons, both with and without a stress
concentration. All the specimens, as described in Chapter 3, were manufactured from
balanced weave 60% WF glass reinforced polypropylene Twintex.

5.2

Analysis Technique

The analysis models of coupon test specimens for this part of the work were based on
the meshes generated during the implicit analysis damage model development. The
use of identical meshes allowed direct comparison of the implicit and explicit codes.

The meshes were generated using HyperMesh 4.0 and exported to the pre-processor,
PAM-GENERIS, where boundary conditions, materials and loads were defined and
applied. All models were defined using reduced integration shell elements. Post
processing was carried out using PAM-VIEW software.

5.2.1 Finite Element Code


The explicit finite element code, PAM-CRASH [81], was selected for this
investigation. The code has various material models available for the simulation of
general non-linear materials as well as more advanced material and damage models
specifically formulated for composites.

5-1

The composite material models in PAM-CRASH allow the user to calibrate behaviour
based on experimental test results although there is no functionality implemented for
adaptation or modification of the models.

5.2.2 Material Model


PAM-CRASH material type 130 was selected for this part of the research programme.
This material type defines a multi-layered shell element with composite options. The
layers can be defined as either elastic damaging fibre-matrix composite or elasticplastic with damage. For the analysis of woven Twintex, the elastic damaging biphase layer definition, ITYP=0, was selected. This damage model is described in
further detail in section 5.3.

As discussed in Chapter 2, this material model has been used with some success to
model the non-linear damaging behaviour of various thermoset matrix composites,
although

little

research

is

available

validating

the

model

for

woven

glass/thermoplastic matrix composites.

5.2.3 Element Definition and Formulation


For all the simulations performed using the PAM-CRASH solver, Belytschko-Tsay
thin shell elements were used [81]. Shells were selected due to the nature of the test
specimens modelled, which all have a small thickness to length ratio. The elements
used to model the composite material test coupons in this study are standard, reduced
integration four noded elements. No three noded elements were used for the coupon
test specimens, although a small number were included in the demonstrator
component models.

Shell elements of this type, defined with material type 130 have one integration point
per composite layer. The total number of through thickness integration points is
therefore dependent on the number of layers in the section definition. For the
simulation of in-plane test specimens, this is not critical since no bending is
simulated, although in later work, when the demonstrator component is deformed out
of plane, simulation of out of plane bending is critical.

5-2

5.2.4 Section Definition


The section definition is referenced from the material type 130 input card. Each layer
of the shell references a ply card, which contains ply material, thickness and
orientation input parameters. For the simulation of coupon test specimens the
composite laminate is defined as a number of layers of nominal 0.5mm thickness.

5.2.5 Definition of Orthotropy Direction


The orthotropy directions for all the woven material models are defined in relation to
the global coordinate system for the model. For shell elements, only the vector of the
principal material direction is required, since the second orthotropy direction is
resolved into the plane of the element. Shear of the fabric during forming and the
resultant variations in fibre direction are beyond the scope of this investigation. For
the geometries analysed here, with little structural complexity, it is assumed that
woven reinforcing fibres are tangential to each other.

5.3

Bi-Phase Damage Model

The bi-phase model allows the user to define the fibre properties of a composite as a
one-dimensional material phase, separately from the properties of the matrix, which
are defined as an orthotropic material. The stiffness of the resulting composite is
calculated by superimposing these two material phases. Damage occurs to both the
fibre and matrix according to individual laws.

This type of material model is applicable to unidirectional fibre reinforced composite


plies. It is less suitable for woven materials such as Twintex.

By neglecting the fibre phase in the bi-phase model, the matrix properties can be used
to model the behaviour of an orthotropic composite material. This approach allows
the analyst to effectively model the behaviour of a woven composite material such as
Twintex.

5.3.1 Degenerate Bi-Phase Damage Model


In its degenerate form, the matrix phase of the bi-phase model is used to model the
elastic damaging behaviour of a woven composite material. The model allows

5-3

calibration based on a combination of both equivalent volumetric strain and


equivalent shear strain dependant damage parameters. For models developed during
this study, damage is identified wholly through (deviatoric) equivalent shear strain,
rather than the (direct) volumetric equivalent strain. This assumption is possible due
to the use of the degenerate form of the bi-phase model, where the fibre is dealt with
as part of a homogenous composite material rather than a separate, superimposed onedimensional fibre phase as in the unidirectional bi-phase model. Damage developed
using a purely deviatoric strain based parameter significantly simplifies the calibration
of the damage magnitudes and strains, but still allows the user to tailor the damage
model accurately for the material.

A single damage parameter, d, is used to quantify the level of damage accumulated by


the material during an analysis, based on the deviatoric component of strain. In the
degenerate form of the model, this damage parameter is used to calculate reduced
elastic properties in the material model for areas where damage has been identified.
The parameter is used to reduce all the elastic modulus properties for the composite
and therefore does not decouple damage modes, unlike the model developed in
Chapter 4. The relationship between strain and damage and subsequently damage and
elastic constants is shown in Figure 2.8, in Chapter 2, where the model was first
introduced.

To summarise the behaviour of the model, the damage parameter, d, is set at zero and
remains at this level until an initial strain threshold is reached. The damage level is
then increased linearly in two stages through an intermediate strain level up until a
point where an ultimate damage level is deemed to have been reached. All of the
composite material elastic moduli are reduced proportionally to the parameter, d,
through all the stages of identified damage.

5.4

Calibration

The calibration of this material model requires the input of elastic properties and
damage parameters for an orthotrpic composite ply in both tension and compression.
These, as mentioned in the definition of the degenerate bi-phase model, are input as

5-4

the matrix phase properties and are damaged according to the deviatoric strain based
damage law.

The calibration parameters were taken from manufacturers published data and from
experimental test and are detailed in Chapter 3. The calibrated material cards used for
simulations presented in this chapter are given in Appendix D. Investigation of the
sensitivity of the model to variation in the damage model calibration was undertaken
and results of this sensitivity analysis are presented later in this chapter.

5.4.1 Calibration of Composite Material Elastic Constants


Experimental methods for the determination of the elastic material properties are
described in Chapter 3. The in-plane tensile and compressive modulus are assumed to
be the same for both orthotropy directions, since the material is a balanced weave
reinforced fabric composite. The through thickness modulus for the composite was
input from manufacturers published data. Since the work was developing a modelling
capability for in-plane damage simulation, the through thickness properties were not
critical to the results.

The shear moduli and poisons ration were calibrated from manufacturers published
data, with an adjustment of the shear modulus from 1.2GPa to 1.04GPa, based on
results from the previous implicit damage modelling work.

5.4.2 Calibration Of Composite Material Damage Parameters


The damage parameters for the strain based volumetric damage scheme were
calibrated using an iterative process. Since the parameters are used to describe an
overall damage state rather than specific damage mechanisms it was necessary to
develop a calibration, which was applicable to both the fibre direction and shear
damage behaviour of the material.

A spreadsheet, developed during previous work by Curtis [15], was used to generate
stress/strain curves for the calibrated material model in tension, compression and
shear. This allowed comparison with experimental curves and fine adjustment of the
damage parameters to best represent all three loading situations.

5-5

5.4.3 Simulation of Calibration Coupon Tests


Following the definition of elastic properties and damage parameters, simulations of
the tests used to generate calibration data for the material model were performed. This
was undertaken to confirm that the procedure had captured the material behaviour as
accurately as possible. Throughout this section of the work, simulation results, like
the experimental results in Chapter 3, are presented as nominal stress/strain.

5.4.4 Tensile 0/90 Test Simulation


The 0/90 tensile test model was run to validate the material and damage calibration
for balanced weave Twintex tested in the fibre direction. The mesh of 170 shell
elements used to model the specimen, was identical to the mesh used in the implicit
simulation, see Figure 5.1. Load was applied as a constant velocity nodal
displacement to one end of the specimen. The other end of the specimen was fully
constrained. All boundary conditions were applied to 50mm sections of the specimen
to represent the jaws used during experimental tests.

A cross section was defined centrally in the specimen, to give cross sectional force
output for the simulation. This was used to derive the nominal stress, based on the undeformed cross section, which is compared to experimental results. Output of the
damage parameter values for each element was also requested during the simulation.
This allowed identification of areas where damage is predicted and where failure
occurs. Similar output was requested for all further simulations in this part of the
work.

The stress/strain response of the model, shown in Figure 5.2, generally compares well
to the experimental results. The experimental result is dominated by the elastic
response of the material followed by catastrophic tensile failure. This simulation
behaves in a slightly different manner due to the nature of the damage model, which is
calibrated to represent a number of damage modes. The damage parameters can be
adjusted to tailor the response of the model, although later sensitivity analysis shows
that due to the sensitivity of the model, this calibration was the most accurate for
general representation of the material.

5-6

Figure 5.3 shows the specimen displacement and damage development during the
simulation. Damage does not develop during the initial stages of the test, as the
specimen is deformed elastically. At approximately 1% strain damage starts to
develop throughout the specimen. This builds up until ultimate failure, which occurs
in the specimen at approximately 3% strain. This ultimate failure, where the damage
level in the specimen approaches 1 occurs across the width of the specimen in the
region of the jaws. This mode of failure is similar to that seen during experimental
test.

This simple test demonstrated that the tensile material model for behaviour in the
fibre direction was calibrated to generally reflect the damaging behaviour of Twintex
in fibre direction tension.

5.4.5 Compressive 0/90 Test Simulation


The mesh used for simulation of the compressive test specimen is shown in Figure
5.4. The majority of the model was constrained to represent the jaws used for loading
the specimen. Loading was applied compressively using a nodal velocity boundary
condition.

Figure 5.5, shows the load displacement response for 0/90 Twintex 1:1 in
compression. The results from simulation of this test show good correlation up until
around 1% strain. In the experimental test, this is the point at which out of plane
deformation starts to occur and the gauge used to measure local strain became
detached form the specimen. The finite element simulation continues past this point,
but with a significant reduction in modulus, as large amounts of damage are predicted
in the specimen.

Predicted damage is compared to a failed test specimen in Figure 5.6. It can be seen
that the location and shape of the failure zone is predicted accurately, although the
strain at ultimate failure cannot be confirmed due to the nature of the data acquisition
during the test.

5-7

5.4.6 Tensile +45/-45 Test Simulation


The +45/-45 tensile test specimen model used an identical mesh and boundary
conditions to the 0/90 tensile model. The critical difference for this analysis was
that the material model was oriented at 45 to the axis of load application.

The stress/strain response of the model and experimental results are shown in Figure
5.7. There are two fundamental differences in the response of the finite element model
to the experimental results obtained. The first is the observed modulus of the
specimens. The model shows that the predicted behaviour of the +45/-45 specimen
is considerably stiffer than that observed. Secondly the strain to failure for the model
is lower than the experimentally observed value.

Figure 5.8 shows the specimen displacement and damage development during the test.
Damage is initiated at the early stages of the displacement and builds up throughout
the specimen. Localised damage is only seen in the later stages of the +45/-45
model analysis. The failure of the model results from a rapid growth in damage at
approximately 2.2% strain, which leads to ultimate failure. This failure is observed in
experimental tests, but at considerably higher strain levels.

5.5

Validation

The same range of coupon tests used during the validation of the implicit finite
element damage model, in Chapter 4, were used during the validation of the
calibration parameters for the PAM-CRASH damage model. During this part of the
work no alteration was made to the previously calibrated parameters. The specimens
selected were relatively simple geometries, which in most cases contained a stress
concentration in the form of a circular hole.

5.5.1 Compressive +45/-45 Test Simulation


For this simulation, the compressive test specimen model was identical to that used in
the 0/90 compressive specimen, although the material model in this case was
oriented at 45 to the axis of the test. The shear behaviour of the material was
calibrated from the off-axis tensile test, so the result from this simulation was used
only for validation.

5-8

The stress/strain response from the simulation and test is shown in Figure 5.9. The
simulation, like the off-axis tensile test, shows a stiffer behaviour than experimental
test. In the test, the specimen begins to damage at very low strain levels and shows a
progressive reduction in stiffness. The simulation shows little reduction in stiffness up
to approximately 0.75% strain. After this point, damage development is more rapid in
the simulation than in the test and the stiffness reduces significantly. Between 1.5%
and 2% strain, the simulation predicts high levels of localised damage and the
specimen begins to fail catastrophically.

Figure 5.10 shows the damage build up in the specimen during the simulation. These
images confirm the response observed in the stress strain curves. At 2.0% strain, 5
elements in the central area of the specimen have reached a maximum level of
damage. This damage progresses rapidly through the specimen, which by 2.5% strain
has failed across its whole width.

This result shows a similar behaviour to the tensile of-axis test, highlighting the
difficulty in calibrating a set of damage parameters to accurately model both shear
and tensile fibre damage. Using the bi-phase model, as shear damage develops, the
fibre direction properties are also damaged, resulting in a more brittle failure than that
observed in test, where the matrix fails, but the fibres remain intact and continue to
function in load bearing up to significantly higher strains than predicted in simulation.

5.5.2 Tensile 0/90 Hole in Plate Test Simulation


Figure 5.11, shows the finite element mesh used for the simulation of the 0/90
tensile hole in plate test simulation. The specimen was constrained and simulated
using identical boundary conditions to the plain tensile test specimens.

The stress/strain response from test and simulation is shown in Figure 5.12. The
simulation correlates well with experiment up to approximately 1% strain and
accurately models the elastic behaviour of the specimen. After 1% strain the
simulation predicts catastrophic failure in the specimen, which is not observed in the
experimental test until approximately 2.0% strain.

5-9

Figure 5.13 shows the development of damage during the simulation, compared to the
failed experimental test specimen. As expected, damage is initiated at the edge of the
stress concentration. In the simulation, this damage leads to rapid catastrophic failure,
as the modulus of the elements at the edge of the hole is reduced and load is
transferred to adjacent elements. Total failure is predicted across the width of the
specimen by 1.2% strain. This result is conservative compared to the experiment,
where, although the specimen starts to damage around the stress concentration,
catastrophic failure does not occur until 2.0% strain. This result is similar to the
behaviour of the implicit finite element damage model observed in Chapter 4, where
in tensile fibre direction simulations, localised damage leads to catastrophic failure.

5.5.3 Tensile +45/-45 Hole in Plate Test Simulation


The off-axis tensile hole in plate specimen uses an identical mesh and boundary
conditions to the previous simulation, although the material model is aligned at 45 to
the axis of the test.

Figure 5.14 shows the stress/strain response from simulation and experimental test.
The model predicts that catastrophic failure will occur at approximately 2.0% strain in
the gauge length of the specimen. This is significantly different to the experimentally
observed result. The simulation also predicts that the stiffness of the specimen prior to
failure will be higher than the test. This increased stiffness was seen in previous off
axis simulations and was expected in this analysis. The premature failure was
expected and had also been observed in previous simulations and is a combination of
over prediction of elastic property reduction resulting form shear damage and load
transfer from damaged elements causing rapid failure in adjacent areas of the
specimen.

The development of damage in the specimen is shown in Figure 5.15. The rapid
failure between 1.5% and 2.0% strain in the specimen gauge length can be clearly
observed. Like the fibre direction tensile test with a stress concentrator, as soon as
catastrophic damage has occurred around the stress concentration, load transfers to
adjacent elements and failure occurs rapidly across the width of the specimen. The
shape of the damaged area is significantly different to the fibre direction simulation
and shows failure occurring at 45 to the axis of the test.
5-10

5.5.4 Compressive +45/-45 Hole in Plate Test Simulation


The compressive off axis test specimen model, with a stress concentration, is
simulated with the mesh shown in Figure 5.16. The stress strain response for the test
and simulation is shown in Figure 5.17. As with the other off axis fibre orientation
simulations, failure occurs significantly earlier in the simulation than in test and the
stiffness of the specimen is predicted to be slightly higher as the specimen is
displaced. At 1mm displacement, the test specimen has damaged significantly and the
stiffness is reduced, a phenomenon which is not captured accurately by the model.
When damage is identified in the simulation, failure is not progressive, as it is in the
test, and occurs in two stages with an initial failure at approximately 1.2mm and a
catastrophic collapse of the specimen at 2.2mm.

Figure 5.18 shows the damage evolution during the simulation. Between 1mm and
2mm it can be observed that the first reduction in specimen stiffness is caused by
damage approaching a level of approximately 0.5 in the elements around the stress
concentration. The second, catastrophic failure is shown at 2.6mm where elements all
across the specimen are approaching a maximum damage level. This specimen, like
the previous simulation, also shows damage evolving at 45 to the axis of the test.

5.6

Model Sensitivity

Since the material model for a composite ply requires extensive calibration, a large
proportion of the work undertaken was an assessment of the sensitivity of analysis
results to variation in material properties. Although an initial calibration was carried
out using a spreadsheet and macro to produce calibration curves based on damage
parameters, further adjustment was undertaken during the simulation of the
calibration tests, detailed in section 5.4. Both the 0/90 and +45/-45 tensile test
models were run with variation in a range of selected parameters to ascertain which
situations and damage types were most sensitive to the calibration.

Subsequently an investigation of sensitivity to mesh refinement, using the +45/-45


compressive hole in plate simulation, was performed.

5-11

5.6.1 Sensitivity to Damage Model Parameters


For the investigation of the sensitivity of the model to variation in calibration
parameters, four models were used. The parameters for these models are detailed in
Table 5.1. Initially these models were used to simulate the 0/90 tensile test.
Stress/strain results from these simulations are shown compared against the final
calibration and experimental test in Figure 5.19. It can be observed from this
comparison that the performance of the model in the fibre direction, as expected, is
highly sensitive to the parameters used. The final calibrated model shows the best
agreement with experiment. Both Model 2 and Model 4 use the same ultimate damage
level, of 0.6, which in the case of this simulation is not high enough to fail the
specimen.

Figure 5.20 shows the simulation results from the +45/-45 tensile test, using the
same model calibrations given in Table 5.1. The stress/strain results show that this test
is far less sensitive to damage model calibration than the fibre direction tensile test.
All the models over predict the stiffness of the specimen and reach ultimate failure at
significantly lower strain levels than was observed in experiment. This result confirms
that the final version of the calibrated model, used in the validation simulations and
later demonstrator component models described in Chapter 6, offers the best
compromise between accurate simulation of fibre direction and shear damage.

Unlike the model developed in Chapter 4, this model degrades all elastic properties
based on the level of damage identified. This leads to reduction in fibre direction
properties when shear damage is identified and necessitates careful calibration to
capture all damage modes as accurately as possible.

5.6.2 Sensitivity to Mesh Refinement


The models sensitivity to mesh refinement was investigated using the compressive
+45/-45 hole in plate specimen simulation. Two further meshes were generated, a
coarse mesh shown in Figure 5.21 and a fine mesh shown in Figure 5.22.

Figure 5.23 shows the stress/strain response for the coarse mesh, compared with test
and Figure 5.24 shows the damage development in the specimen. The specimen
predicts first significant damage at slightly over 1mm of displacement and
5-12

catastrophic failure just before 3mm. This result is similar to that seen in section 5.5.4,
although failure is predicted slightly later in the simulation and at a higher load.

The stress/strain results and damage plots for the fine mesh, shown in Figure 5.25 and
Figure 5.26 respectively show a slight difference in results to the coarse mesh. Failure
occurs earlier and at a lower load. From the stress/strain response, it can be seen that
the first major damage occurs at just before 1mm displacement and ultimate failure
occurs at just before 2mm. These results are very similar to those from the standard
mesh simulation performed during the validation phase of the work. This suggests
that the damage prediction has converged, as there is little sensitivity or improvement
in accuracy of the simulation between the standard and the fine mesh.

5.7

Conclusion

The PAM-CRASH degenerate bi-phase damage model for composite materials has
been used to simulate the damaging behaviour of balanced weave Twintex. The
model has been calibrated using a range of data from experimental test and
manufacturers datasheets.

Calibration tests and a range of validation tests have been simulated to investigate the
models performance in a range of in-plane damage situations. This has shown that
the model accurately predicts the behaviour of Twintex damaging in the fibre
direction. Shear damage in the specimen has been shown to be less accurately
modelled by the technique, with an over prediction of the stiffness of all the off-axis
specimens simulated.

Specimens with a hole showed that the model was also sensitive to geometric
features. As damage started to develop in the region of the stress concentrator the
model often predicted rapid degradation of material properties and lead to
catastrophic failure. The sensitivity of the model to the calibration parameters was
significant when modelling fibre direction damage. This sensitivity was not observed
in the simulation of shear damage.

5-13

5.8

Tables

Parameter
iT
1T
uT
d1T
duT
iC
1C
uC
d1C
duC
Table 5.1

Model 1
0.007
0.02
0.05
0.2
0.8
0.007
0.02
0.05
0.2
0.8

Model 2
0.01
0.02
0.08
0.1
0.6
0.01
0.02
0.08
0.1
0.6

Model 3
0.0053
0.017
0.039
0.5
0.95
0.0053
0.017
0.039
0.5
0.95

Model 4
0.0053
0.017
0.039
0.1
0.6
0.0053
0.017
0.039
0.1
0.6

Material damage model parameters for sensitivity analysis

5-14

5.9

Figures

Figure 5.1

Tensile test - shell element mesh

Figure 5.4

Compressive test - shell element mesh

Figure 5.11

Tensile hole in plate test - shell element mesh

Figure 5.16

Compressive hole in plate test - shell element mesh

Figure 5.21

Compressive hole in plate test - shell element mesh (coarse)

Figure 5.22

Compressive hole in plate test - shell element mesh (refined)

5-15

300

Experimental Test
PAM Crash

250

Stress (MPa)

200

150

100

50

0
0

0.5

1.5

2.5

Strain (%)

Figure 5.2

0/90 tensile test simulation - stress/strain

0.67%, 1.33%, 2.0%, 2.67%, 3.0%

Figure 5.3

0/90 tensile test simulation - experimental and predicted damage

5-16

3.5

150

Experimental Test
PAM Crash

Stress (MPa)

100

50

0
0

0.5

1.5

2.5

3.5

Strain (%)

Figure 5.5

0/90 compressive test simulation - stress/strain

1.0%, 2.0%, 3.0%, 3.1%

Figure 5.6
damage

0/90 compressive test simulation - experimental and predicted

5-17

60

Experimental Test
PAM Crash

50

Stress (MPa)

40

30

20

10

0
0

0.5

1.5

2.5

3.5

Strain (%)

Figure 5.7

+45/-45 tensile test simulation - stress/strain

0.67%, 1.33%, 2.0%, 2.13%, 2.27%

Figure 5.8

+45/-45 tensile test simulation - experimental and predicted damage

5-18

45

Experimental Test
40

PAM Crash

35

Stress (MPa)

30
25
20
15
10
5
0
0

Strain (%)

Figure 5.9

+45/-45 compressive test simulation - stress/strain

0.5%, 1.0%, 1.5%, 2.0%, 2.5%, 3.0%

Figure 5.10
damage

+45/-45 compressive test simulation - experimental and predicted

5-19

200

Experimental Test

180

PAM Crash

160

Stress (MPa)

140
120
100
80
60
40
20
0
0

0.5

1.5

2.5

Strain (%)

Figure 5.12

0/90 tensile hole in plate test simulation - stress/strain

0.5%, 0.7%, 0.8%, 1.0%, 1.2%, 1.4%

Figure 5.13
damage

0/90 tensile hole in plate test simulation - experimental and predicted

5-20

100
Experimental Test

90

PAM Crash

80

Stress (MPa)

70
60
50
40
30
20
10
0
0

10

12

14

Strain (%)

Figure 5.14

+45/-45 tensile hole in plate test simulation - stress/strain

1.0%, 1.5%, 2.0%, 4.0%, 6.0%

Figure 5.15 +45/-45 tensile hole in plate test simulation - experimental and
predicted damage

5-21

Experimental Test
3.5

PAM Crash

Load (kN)

2.5
2
1.5
1
0.5
0
0

0.5

1.5

2.5

3.5

4.5

Displacement (mm)

Figure 5.17 +45/-45 compressive hole in plate test simulation - load/


displacement
1mm, 2mm, 2.4mm, 2.6mm, 3mm

Figure 5.18 +45/-45 compressive hole in plate test simulation - experimental and
predicted damage

5-22

700

Experimental Test
Model 1
Model 2
Calibrated Model
Model 3
Model 4

600

Stress (MPa)

500
400
300
200
100
0
0

10

Strain (%)

Figure 5.19

0/90 tensile test simulation sensitivity

80

Experimental Test
Model 1
Model 2
Calibrated Model
Model 3
Model 4

70

Stress (MPa)

60
50
40
30
20
10
0
0

2
Strain (%)

Figure 5.20

+45/-45 tensile test simulation sensitivity

5-23

Experimental Test
3.5

PAM Crash

Load (kN)

2.5
2
1.5
1
0.5
0
0

0.5

1.5

2.5

3.5

4.5

Displacement (mm)

Figure 5.23

+45/-45 compressive hole in plate test simulation - coarse mesh

1mm, 2mm, 3mm

Figure 5.24 +45/-45 compressive hole in plate test simulation - coarse mesh predicted damage

5-24

Experimental Test
3.5

PAM Crash

Load (kN)

2.5
2
1.5
1
0.5
0
0

0.5

1.5

2.5

3.5

4.5

Displacement (mm)

Figure 5.25

+45/-45 compressive hole in plate test simulation - fine mesh

1mm, 2mm, 2.4mm, 2.6mm, 3mm

Figure 5.26 +45/-45 compressive hole in plate test simulation - fine mesh predicted damage
5-25

Chapter 6

Application to an Automotive Demonstrator

Component
6.1

Introduction

To validate the manufacturing methods and analysis techniques and to compare the
performance of a thermoplastic part with a metallic equivalent, an automotive
demonstrator component programme was undertaken. Various factors influenced the
choice of demonstrator component, including size, geometry and cost of manufacture
and test, as well as the potential viability of introducing the part on future vehicles.
For these reasons, a side intrusion protection beam was selected.

Initially, at the start of the research programme, a Phase 1 demonstrator component


was designed. At this stage, analysis capability was limited and unvalidated, so the
design was based on a simple section geometry designed to offer an acceptable level
of stiffness. This beam was thoroughly tested and results were used to validate the
finite element damage modelling techniques developed during the core part of the
research programme.

The Phase 2 demonstrator component was developed with the aim of installation and
test in a full vehicle, to an accepted legislative test method. During the design of the
beam, the validated finite element damage models were used extensively to
investigate potential geometries and select the most effective. This process reflects the
design and simulation stages undertaken during commercial projects developing new
components. The Phase 2 demonstrator component was tested and results were
compared to those from fully predictive finite element simulations. This testing
evaluated the performance of the beam against a current design and further validated
the explicit finite element damage modelling technique developed and calibrated in
Chapter 5.

For the final full vehicle test a Phase 2 beam was installed in one side of a vehicle
along with a thermoplastic door cassette component. Both sides of the vehicle were
tested to the Federal Motor Vehicle Safety Standard for side intrusion performance

6-1

[45] and results for the composite concept side and the standard steel door
configuration were compared.

During the development of the demonstrator component a further two forms of


Twintex were also investigated. These were 4:1 plain weave, 60% weight fraction,
commingled glass polypropylene Twintex and a double layer, balanced, 2 x 2 twill
weave, 60% weight fraction, commingled glass polypropylene with 1% z-axis glass
fibres, in this study referred to as 3D Twintex.

6.2

Side Intrusion Beam Test Methods

The Phase 1 demonstrator component, described in section 6.3, was tested using two
different 3-point bending test methods, referred to as a small and large-scale 3-point
flexure test.

The Phase 2 demonstrator component, described in section 6.4, was tested using the
large-scale 3-point flexure test method and was validated in a test vehicle, using the
FMVSS 214 [45] quasi-static pole side intrusion test.

6.2.1 Small-Scale 3-Point Flexure Test


For preliminary testing of the Phase 1 beam a 3-point flexure rig was developed for a
servo-hydraulic Instron 8500 test machine. Due to the dimensions of the test machine
and fixture, the maximum deflection of the beam was limited to 80mm. The load was
applied at the middle of the beam through a 2 inch diameter cylindrical impactor and
each end of the beam was supported with a 2 inch diameter cylindrical bar.

Aluminium brackets, manufactured from sheet material, were used to mount an


extensometer across the centre of beam for measurement of cross-sectional splaying.
Measurement of splaying was made up to 5mm during the test. A Phase 1 beam
mounted in the test rig is shown in Figure 6.1.

Load data was measured using a 100kN load cell and deflection data was taken from
the crosshead of the test machine. The beams were tested at a crosshead displacement

6-2

rate of 10mm/minute. Table 6.1 details the thickness, material and lay-up of the Phase
1 beams tested in the small-scale 3-point flexure rig.

6.2.2 Large-Scale 3-Point Flexure Test


To test the beams past the early onset of damage and in the cases of the aligned fibre
beams, to ultimate failure, a larger scale 3-point flexure test was developed. The rig
used was based on an FMVSS 214 test rig, set up with the beam mounted rigidly at
one end and with a compliant steel bracket and restraining chains at the other end. The
rig, during Phase 1 beam testing is shown in side view in Figure 6.2 and from the
front in Figure 6.3. Figure 6.4 shows the rig during testing of the Phase 2
demonstrator component. A standard steel side intrusion beam from the target vehicle
was also tested during the Phase 2 programme to allow comparison with the
demonstrator component.

Load data was measured using a load cell mounted behind the 12 inch diameter
impactor and displacement was measured using a wire pull displacement transducer
attached to the impactor. Both the Phase 1 and Phase 2 beams were tested at a
nominal displacement rate of 500mm/minute.

Table 6.2 details the thickness, material and lay-up of the Phase 1 and Phase 2 beams
tested in the large-scale 3-point flexure rig.

6.2.3 FMVSS214 Side Intrusion Test


The Phase 2 demonstrator was designed to meet specific load and displacement
requirements detailed in FMVSS214, the federal Motor vehicle Safety Standard for
side intrusion performance [45]. As part of a wider research programme a semistructural door cassette was investigated, so this was also used in the final validation
test. The door interior panel and steel beam were removed and the composite beam
was mounted in the door using steel brackets. The composite door cassette,
manufactured from glass mat thermoplastic (GMT) was mounted in place of the
standard door trim panel. The assembled door mounted in the test vehicle is shown in
Figure 6.5.

6-3

The vehicle was constrained by bolting the chassis to a hard floor. The 12 inch
diameter impactor used in the large-scale 3-point flexure tests was used to quasistatically deform the doors centrally, see Figure 6.6. The vehicle was tested at a
nominal displacement rate of 500mm/minute. Load displacement data was measured
during the test using the same equipment as that used in the large-scale 3-point flexure
tests. Table 6.3 details the thickness, material and lay-up of the Phase 2 beam and
steel beam tested to FMVSS214.

6.3

Phase 1 Demonstrator Component

The Phase 1 side intrusion beam demonstrator component was developed during the
early stages of the project. It was used to validate both the manufacturing and
modelling techniques being developed during the research programme. Since
modelling tools were not fully developed at the time of the initial design, only basic
numerical modelling was undertaken to evaluate the initial stiffness of the component.

A test programme, including small and large-scale flexural testing was undertaken.
Results from this testing were compared to simulations performed subsequent to the
tests.

6.3.1 Phase 1 Demonstrator Component Design


The component was developed to a generic design envelope, similar to that available
for current steel beams, see Figure 6.7, which would allow the beam to be mounted in
a typical 3 door vehicle. This approach set constraints on both the length and depth of
the beam.

Since the beam was to be tested in flexure, compressing one surface and loading the
other in tension, the compressive and tensile strength of Twintex was taken into
account. For this reason, the final geometry was developed as a top hat section, with
approximately double the quantity of material on the compressive face, to take into
account the reduction in strength observed in compression during coupon tests. The
final design is shown in Figure 6.8 and a section through the centre of the beam is
shown in Figure 6.9

6-4

6.3.2 Phase 1 Demonstrator Component Manufacture


The beam was manufactured using matched metal tooling and a non-isothermal
compression moulding process. The Phase 1 demonstrator component tool included
both a sprung blankholder and shear edge, based on the flat plaque tool concept
introduced in Chapter 3.

The processing conditions for the demonstrator components were less controllable
than those used in the laboratory, due to the industrial scale techniques employed,
although where possible, the previously specified optimised parameters, given in
Chapter 3, were used.

6.3.3 Phase 1 Demonstrator Component Small-Scale 3-Point Flexure Test


Two 3-point flexure test methods, described in section 6.2, were used to evaluate the
performance of the Phase 1 demonstrator component. The first method, a small-scale
3-point flexure test, was performed on four beams: one of each of the four material
configurations moulded. The load applied to displace the beam was measured as well
as the splaying of the section, since the deformation mode was expected to vary
depending on the material configuration.

Figures 6.10 and 6.11 show the results from the 0/90 Twintex beam. The loading is
linear during the first stages of the displacement, but starts to show a reduction in
stiffness as the beam displaces and the section splays. The first critical failure in the
beam occurs in the corner of the lower flange at a displacement of approximately
63mm and a load of 4.7kN. This area continues to damage up until the point at which
the test is stopped. No other areas of the beam showed significant levels of visible
damage up to this point. The beam had splayed 5mm, the maximum measurable using
the extensometer, at a displacement of 47mm.

Figures 6.12 and 6.13 show the results from the +45/-45 Twintex beam. This beam
is less stiff than the 0/90 beam, but displays a similar behaviour, with a reduction of
stiffness during loading. The first critical failure occurs at 72mm and a load of 2.8kN,
again in the lower flange corner of the beam. The failure was not as marked as that
seen in the previous beam and no significant cracking was visible. The splaying of the

6-5

section during this test was also significantly lower, with the maximum measurable
limit, of 5mm, being reached at a vertical displacement of 65mm.

Figures 6.14 and 6.15 show the results from the 4:1 Twintex beam. As expected, this
beam was significantly stiffer than the balanced weave Twintex beam. The first
critical failure, in the lower flange corner, occurs at 47mm displacement and a load of
5.4kN. The beam continues to damage in this area up to the end of the test, with no
visible damage being observed in any other location. The beam has less transverse
fibres than the others tested and exhibits the largest amount of splaying during
displacement, with the maximum 5mm being reached at a vertical displacement of
only 35mm.

Figures 6.16 and 6.17 show the results from the 3D Twintex beam. This beam
exhibits a similar stiffness to the 4:1 Twintex beam, although first critical failure in
the flange corner occurred at a higher displacement and load of 67mm and 6.2kN
respectively. The beam showed a similar level of section splaying as the 0/90
Twintex beam, with the maximum 5mm being measured at a displacement of 55mm.
Like the previous beams tested, no visible damage was observed in other locations on
the beam.

In terms of overall stiffness and integrity of the section, the 3D Twintex beam
outperformed all other configurations, with a higher first failure load and
displacement and a reduced amount of cross sectional splaying.

6.3.4 Phase 1 Demonstrator Component Large-Scale 3-Point Flexure Test


The second, large-scale 3-point flexure, test was performed on two beams of each of
the four configurations.

Figure 6.18 shows the result from the large-scale flexure test on the two 0/90
Twintex beams. The results initially show a similar behaviour, with a steady increase
in load up to a first failure point. After this point, the load rises again and leads to
catastrophic failure in the first beam at 225mm displacement. The second beam shows
a drop in load at this point, but does not fail and continues to load up to ultimate

6-6

failure at 295mm displacement. It was observed that the stiffness of the first beam
was significantly higher than the second beam.

Figure 6.19 shows the result from the two +45/-45 Twintex beams tested. The
beams in this test showed a much more progressive loading behaviour, with shear
damage gradually building up during the test and very little catastrophic failure in the
beams. Beam 2 did show a measurable drop in load at a displacement of 270mm
although this did not lead to ultimate failure. After the event the beam continued to
perform as previously, suggesting that the load drop was a phenomenon introduced by
slip in the test rig rather than failure in the beam. Both beams continued to carry load
without failure up to the 305mm maximum displacement in the test. During these
tests, the performance of the two beams was almost identical, when compared to the
variation seen during the test of the 0/90 beams.

Figure 6.20 shows the result from the two 4:1 0/90 Twintex beams. Both beams
exhibited similar behaviour, with an initial failure at 60mm displacement followed by
reloading up to secondary failure at 200mm. Ultimate failure occurred only in Beam 2
at 280mm displacement. Beam 1 did not fail during the 305mm test displacement.
The initial stiffness of both beams was very similar although the first failure in Beam
2 appeared more significant with a load drop of 50% which was larger than that
observed in Beam 1. Initial failure in both beams occurred at a load of approximately
7kN, which is similar to the initial failure load observed in the 0/90 Twintex beams.

The results from the 3D Twintex beam tests are shown in Figure 6.21. These beams,
again, exhibited an initial failure, secondary failure and final failure behaviour. The
initial failure for these beams, between 7.5kN and 8kN was slightly higher than that
observed in the other tests and occurred in both beams at approximately 70mm
displacement. The secondary failure also occurred at a slightly higher load, 10.5kN,
and resulted in only a slight drop off in load. Beam 2 was the only beam to fail during
the 305mm of the test, at 295mm displacement. The ultimate load for the failed beam
was 22.5kN, a significantly higher load than observed in either the failed 1:1 0/90
and 4:1 0/90 beams.

6-7

Two distinct damage patterns were observed during this stage of the testing. The 1:1
0/90 Twintex, 4:1 0/90 Twintex and the 3D Twintex beams, with fibres aligned
along the length of the beam, behaved in a similar manner. Initially, as the beams
were loaded, bending deformation occurred up until approximately 50mm
displacement. At this point a visible and audible damage event occurred, in the lower
flange corner in the middle of the beam. This damage zone was similar to that
observed during the small scale flexure testing. As the beams continued to displace, a
crack progressed up the sidewall of the beams from the damage initiation point and
the section began to flatten. By approximately 200mm displacement, the damage had
progressed to the upper corners of the beam, adjacent to the top face. At this point a
secondary damage zone began to form with a crack in the lower flange progressing up
the sidewall approximately 100mm further away from the built-in end of the beam.

The +45/-45 Twintex beams showed a different damage behaviour to the other
beams tested. Damage was developed more progressively in the beams, with no
visible cracking on the surface of the beams. Shear damage could be observed during
the test, particularly in the sidewall of the beams and splaying of the section was also
seen.

Like the results from the small-scale 3-point flexure tests, the larger scale flexure tests
showed a significant performance improvement in the 3D Twintex beam compared to
the other configurations. These tests showed higher stiffness and failure loads, in
addition to an apparent higher damage tolerance.

6.3.5 Phase 1 Demonstrator Implicit Finite Element Damage Modelling


The damage model developed and calibrated for the ABAQUS implicit finite element
code, see Chapter 4, was used to simulate the tests performed on both the balanced
weave 0/90 Twintex and the +45/-45 Twintex beams. The 4:1 and 3D Twintex
beam tests were not simulated since a damage model was not calibrated for these
configurations, although the model presented in Chapter 4 could be calibrated and
used for these materials.

A single layer shell element model, containing 700 elements with a nominal edge
length of 15mm, was generated to represent the beam geometry and rigid cylinders
6-8

were used to apply a prescribed displacement to the beams. The edge length was
selected as the minimum necessary to allow and accurate description of the geometry
of the beam. Figure 6.22 shows the mesh and boundary conditions for the small-scale
3-point flexure test models and Figure 6.23 shows the mesh and boundary conditions
used for the large-scale test models. In the large-scale test, one end of the beam was
fully constrained and the other was constrained in the direction of the load
application, but free to move in the longitudinal direction of the beam. In both cases,
the load and displacement during the simulation were measured at the rigid wall used
to apply the loading.

Figure 6.24 shows the simulation result for the 0/90 Twintex small-scale flexure
test, compared to the experimental result. The simulation shows good correlation to
test, with accurate prediction of the stiffness of the composite beam. The first failure
point is also predicted at the correct load and displacement, although the simulation
overestimates the amount of damage and predicts catastrophic failure in the beam. In
the test the first failure was not catastrophic and the beam retained structural integrity
and load carrying capability. The damage contour plots shown in Figure 6.25 identify
compressive failure in the lower flange of the beam, initiating in the corner of the top
hat section at around 60mm displacement and progressing rapidly across the beam.
The contour plots of FV3 (field variable 3) representing shear damage show that the
beam damages due to shear as well as fibre direction stress. This shear damage is not
catastrophic and only results in a slight reduction in the beams performance.

Figure 6.26 shows the simulation and experimental result for the +45/-45 Twintex
beam, small-scale flexure test. In this case, the model slightly under-predicts the
stiffness of the beam and does not accurately capture the drop off in load seen during
the experimental test. This simulation result, showing a lower stiffness, is caused by
excessive shear damage development in the model. Figure 6.27 shows FV3 during the
simulation, with the maximum amount of shear damage being predicted in the lower
flange and a minimal amount predicted in the side webs.

The simulation of the large-scale flexure test for the 0/90 Twintex beam is shown in
Figure 6.28. The ABAQUS damage model in this case predicts catastrophic failure in
the early stages of the test, at approximately 50mm. The stiffness predicted by the
6-9

model is similar to that of Beam 1 although it is higher than Beam 2. The damage
contour plots shown in Figure 6.29 identify the ultimate failure in the beam being
predicted as a compressive failure in the lower corners of the beam, identified by FV1
reaching a value of 1 at 50mm of displacement in this area. This result is similar to
that seen in the small-scale test. The FV3 plots show that a small amount of shear
damage is developed during the simulation, but not enough to cause a significant drop
in performance.

Figure 6.30 shows the result from the simulation of the large-scale flexure test on the
+45/-45 Twintex beams. The comparison between analysis and experimental results
shows that the damage model over-predicts the magnitude of shear degradation in the
specimen. The analysis curve shows that the beam is predicted to be significantly less
stiff than the physical specimen. Figure 6.31 shows the damage in the specimen
during simulation. As expected, shear damage is predicted to occur throughout the
central section of the beam and develop rapidly throughout the analysis. There is a
small amount of fibre direction damage predicted to occur between 140mm and
200mm displacement, in both the central area of the beam and in the region of the
sliding constraint. This damage corresponds to the drop off in load observed at
160mm in the predicted force displacement curve.

6.3.6 Phase 1 Demonstrator Explicit Finite Element Damage Modelling


The damage model, calibrated for the PAM-CRASH explicit finite element code, in
Chapter 5 was also used to simulate the balanced weave Twintex demonstrator
component tests. The same shell element mesh and boundary conditions, shown in
Figures 6.22 and 6.23, used for the implicit finite element models, were also used for
the explicit simulation.

Figure 6.32 shows the result from the small-scale flexure test on the 0/90 Twintex
beam. The model accurately predicts the stiffness of the beam, but slightly
underestimates the damage developed at around 50mm, where the test shows a load
drop off which is not observed in the analysis result. The experimental curve does not
show a catastrophic failure in the beam and it is therefore not possible to accurately
extrapolate the experimental curve to compare it to the models behaviour between
70mm and 100mm. Figure 6.33 shows the predicted damage in the beam. It is
6-10

observed that damage is predicted to occur in a fairly uniform manner throughout the
top flange of the beam. In the central section of the beam, the maximum level of
damage, 0.5, is observed in the corner of the lower flange. This corresponds to the
areas of maximum damage observed in the test specimens.

Figure 6.34 shows the result of the simulation of small-scale test on the +45/-45
beam. The correlation between analysis and experiment is good, although the
simulation slightly under predicts the stiffness of the beam. The plots of damage
development during the simulation, shown in Figure 6.35, show that significantly
more damage is predicted in this simulation than the previous analysis. The area of
maximum damage is also spread more uniformly along the top flange of the beam.

The simulation of the large-scale flexure tests on the Phase 1 demonstrator component
show much better correlation throughout the test than the implicit model. The ability
of the explicit code and damage model to cope with large deformations and high
levels of damage is highlighted by these simulations. Figure 6.36 shows the
comparison between experiment and simulation for the 0/90 beam. The prediction
accurately captures the stiffness of the beam and the two major failure points at 75mm
and 225mm. The damage plots shown in Figure 6.37 identify that the first failure
point is caused by material failure in the centre of the beam at the lower corner and
the second major failure is a combination of cracking along the lower radius of the
beam and failure at the sliding constraint.

Figure 6.38 shows the result from the simulation of the +45/-45 Twintex beam
simulation. The analysis slightly over predicts the stiffness of the beam during the
early stage of the analysis, up to 100mm, where a significant failure is predicted. The
beam is then predicted to deform and damage progressively, mirroring the result from
both experimental tests. The damage plots given in Figure 6.39 show that the first
failure predicted is due to localised damage in the centre of the beam. The gradual
development of damage is then observed, without the catastrophic failure seen in the
0/90 beam simulation. The final damage plot, at 280mm, shows that damage has
developed along the top face of the beam. The total failure at the constraint observed
in the previous simulation does not occur during this analysis, although the damage
levels in this area are significant, approximately 0.75.
6-11

6.4

Phase 2 Demonstrator Component

The Phase 2 demonstrator component was developed, based on the Phase 1


component, specifically for a current model, production vehicle. This component
allowed the damage modelling techniques developed and validated during the
research, to be assessed as a design tool for composite materials.

Since the beam was to be fitted to a vehicle, packaging constraints led to a modified
space envelope compared to the Phase 1 beam. Various geometries were investigated
before the final design was selected.

The beam was tested using the large-scale 3-point flexure test method developed for
the Phase 1 component and results were compared to predictive simulations. A
component was then selected for test in the target vehicle. This test was used to
compare the performance of a composite component against the current steel beam
used in production and to validate the design.

6.4.1 Phase 2 Demonstrator Component Design


The validated explicit finite element damage model, detailed in Chapter 5, was used
extensively during the design of the Phase 2 demonstrator component. Initially a
number of geometries were modelled to assess the stiffness of various alternatives.
The four geometries identified as potentially offering an acceptable level of
performance are shown in Figure 6.40. This design matrix compares first failure load,
displacement and mass. This predictive work was used to identify a design to be
developed for final testing and installation in the target vehicle.

The results show that Concept 4 is significantly stiffer than the other designs due to
the constant section running the length of the beam. This is different to the other
concepts investigated, which all taper to a flat section at either end. This design
offered the best compromise between stiffness and first failure displacement, although
required extra design and development work to mount in the target vehicle.

Figure 6.41 shows the final design of the Phase 2 demonstrator component. A section
through the beam is given in Figure 6.42. This shows that like the Phase 1 beam, the

6-12

Phase 2 design has a larger area on the compressive face due to the variation in
strength when comparing the tensile and compressive performance of Twintex. Since
the beam was developed for installation in a specific target vehicle, the CAD
geometry was developed as part of a larger model of the vehicle door. Figure 6.43
shows the beam virtually installed in the door and the clearance between the beam
and window glass, confirming that packaging requirements are met. Subsequently,
mounting brackets were designed to install the beam in the vehicle. Addition of the
brackets caused a small interference in the target door, leading to modification of the
beam geometry at one end. This adjustment to the geometry is shown in Figure 6.44
and the steel brackets are shown in Figure 6.45.

6.4.2 Phase 2 Demonstrator Component Manufacture


The Phase 2 demonstrator component used a simpler tool concept, a matched metal
tool without a shear edge or blankholder. This second demonstrator component tool
was used to investigate the use of inserts to vary thickness without geometry
inaccuracy in the radius areas. The beam was initially developed as an 8mm thick
section, but was also moulded with 6mm and 4mm sections, through the use of two
2mm thick tool inserts. This allowed further component test data to be generated for
comparison with analytical models.

The tool is shown in Figure 6.46, with a 2mm thick insert installed. The inserts were
formed from 2mm sheet steel, by compression in the tool. Figure 6.47 shows a closer
view of the insert, in the most complex area of the tool, around the step in geometry.
A detailed description of the non-isothermal compression moulding process and
optimised parameters are given in Chapter 3.

Figure 6.48 shows the industrial moulding process used for the Phase 2 demonstrator
component. The process is identical to the laboratory scale plaque moulding technique
presented in Chapter 3. Preconsolidated blanks are heated in an infrared oven and
transferred to a cool press tool. The component is then formed and consolidated as it
cools, before removal from the mould. This process was used for all three forms of
Twintex used in the manufacture of demonstrator components.

6-13

6.4.3 Phase 2 Demonstrator Component Large-Scale 3-Point Flexure Test


The Phase 2 demonstrator component was tested using the large-scale 3-point flexure
test method, employed in the second stage of the testing of the Phase 1 beam. All
beams were tested to a nominal 305mm displacement with load data being measured
at the impactor. Two beams each of 4mm, 6mm and 8mm thickness, manufactured
from 1:1 0/90, 1:1 +45/-45 and 4:1 0/90 Twintex as well as two 6mm 3D 0/90
Twintex beams were tested. Typical tested specimens of each of the four material
configurations are shown in Figures 6.49 to 6.52. These tested specimens showed
characteristic failure types for each material and are compared to predicted damage in
section 6.4.4.

Figures 6.53, 6.54 and 6.55 show the results from the three thickness variants of the
0/90 1: 1 Twintex beams tested. All six beams showed a similar displacement
behaviour, with a linear loading up to the first failure point, a drop in load and then
progressive rise up to a second failure point. The 4mm beams did not exhibit a
catastrophic failure within the 305mm of test, although beam 2 showed a major failure
at 280mm with an 80% drop in load carrying capacity. The 6mm beams exhibited a
similar stiffness to the 4mm beams and a first failure point at a significantly higher
displacement. This result was unexpected as the extra section thickness was predicted
to increase the stiffness of the component over that of the 4mm beams. The 8mm
beams did show a significant stiffness increase when compared to the previous
results. The first failure point for these beams occurred at approximately 50mm
displacement, similar to the 4mm beams. This suggested that the 6mm beam results
were uncharacteristically poor when compared to both the 4mm and 8mm beams.

Figures 6.56, 6.57 and 6.58 show the results from the three thickness variants of the
+45/-45 1: 1 Twintex beams tested. The beams all showed a similar progressive
damaging behaviour, with no rapid failure or drop in load carrying capacity. Unlike
the 0/90 beams, the step up in thickness from 4mm to 6mm resulted in a significant
increase in stiffness of the beams, as did the step from 6mm to 8mm. The
performance when compared for the two beams of each configuration, showed
variability in the 6mm and 8mm beams, but a virtually identical performance up to
200mm displacement, for the 4mm beams.

6-14

The results from the 4:1 Twintex beams tested are shown in Figures 6.59, 6.60 and
6.61. These tests show a similar result to the 0/90 1:1 Twintex beams, with a
predominantly linear loading behaviour up to the first failure point followed by a drop
in load carrying capacity and then a progressive reloading. Both the 8mm beams
failed catastrophically within the 305mm of test displacement, at approximately
285mm. This type of failure was not observed in the thinner sectioned beams, where
the load progressively increased up to 305mm. Like the +45/-45 Twintex beams, the
4:1 configuration showed an increase in performance between the 4mm, 6mm and
8mm beams respectively.

The final material configuration tested was the 0/90 3D Twintex. These beams were
only tested in the 6mm configuration. The results from the two tests are shown in
Figure 6.62. The load displacement behaviour shows the characteristic initial linear
loading followed by failure and reloading. The beams both exhibited catastrophic
failure at 280mm, a result not seen in other material configuration 6mm beam tests.
The performance of the beams, in general, was significantly better than the standard
0/90 1:1 Twintex in terms of stiffness and initial failure load. It was also observed
that the results showed a consistent behaviour between the two beams tested.

Like the Phase 1 beam tests, two distinct damage behaviours were observed for the
Phase 2 beams. The components with fibres aligned along the length beam showed
localised damage and failure, whereas the +45/-45 Twintex beams all showed a
progressive shear damaging behaviour in the sidewalls with some splaying of the
cross-section at higher displacements. The 0/90 fibre orientation beams, all loaded
linearly up until approximately 50mm where the first localised damage occurred in
the middle of the beam in the central section of the lower flange. Soon after this event,
secondary areas of damage occurred in the outer corners of the lower flange. Cracks
then propagated up the sidewalls of the section, as the beam was displaced up to
approximately 220mm. At this point, a third damage zone developed approximately
150mm further away from the built in end of the beam. This again produced cracks
running up the side wall of the beam, which lead to ultimate failure in the case of the
3D Twintex beams.

6-15

6.4.4 Phase 2 Demonstrator Implicit Finite Element Damage Modelling


Although simulation of the Phase 1 demonstrator had shown that the implicit finite
element damage model was not ideally suited to large displacement simulations, it
was still used to simulate the behaviour of the tested beams. All results presented in
this section were produced before the Phase 2 beams had been moulded or tested and
are therefore, truly predictive simulations. Since the model was developed and
calibrated for 1:1 Twintex, only the 0/90 and +45/-45 1:1 Twintex beams were
simulated. All three thicknesses were modelled using the properties of the shell
elements to simulate the 4mm, 6mm and 8mm variants. The model, containing 1880
shell elements with a nominal edge length of 10mm, and the rigid wall boundary
condition for the simulations are shown in Figures 6.63 and 6.64. One end of the
beam was constrained rigidly using a nodal constraint and the other was constrained
to move only in the longitudinal direction of the beam. These conditions represented
the test fixture used for the large-scale 3-point flexure test. The nominal element edge
length of 10mm was selected to represent the typical shell element dimensions seen in
an automotive finite element crash simulation model.

Figure 6.65 shows the implicit finite element damage model simulation results for the
4mm thick 0/90 Twintex beam. The load displacement result shows that the damage
model predicts damage earlier in the displacement than was observed during test. This
failure results in a rapid reduction in load carrying capacity and leads to a failure to
converge a solution at approximately 90mm. The model also slightly over-predicts the
stiffness of the test components during the early stages of the test, from 0mm to
25mm. The damage contour plots, in Figure 6.66, show that the load drop is caused
by compressive failure in the lower areas of the beam section. A small amount of
shear damage is predicted in the beam, although this does not significantly reduce the
structural performance when compared to the level of fibre direction damage
identified.

Figure 6.67 shows the predicted result for the 6mm thick 0/90 Twintex beam. The
stiffness of the response from the analysis model is significantly higher than that
observed during test and like the 4mm beam, failure is predicted in the early stages of
the simulation. Figure 6.68 shows that this drop in load is predominantly the result of
compressive, fibre direction, failure in the lower flange of the beam. Small amounts of
6-16

shear damage are identified in the beam. Again, this does not significantly affect the
performance.

The results from the 8mm 0/90 Twintex beam simulation are shown in Figure 6.69.
This analysis, when compared to all the 0/90 beam simulations, gave the most
accurate prediction of beam stiffness, although like previous results, showed a
tendency for over-prediction of damage. Failure was predicted at 30mm displacement
compared to the experimentally observed 50mm displacement at first failure. The
predicted fibre direction damage again lead to a failure to converge a solution, due to
deformation in elements in which the modulus had been significantly reduced. This
damage is shown in the contour plots given in Figure 6.70. The failure occurs on the
compressive face of the beam, but in a more localised area than the previous two
simulations.

Figure 6.71 shows the result from the simulation of the 4mm +45/-45 Twintex
beam. The load displacement behaviour and damage development during the early
stages of the test are predicted accurately by the damage model. The ultimate failure,
predicted at 120mm is inaccurate, since this was not observed in the test. The plot of
damage contours shown in Figure 6.72 shows that shear damage builds progressively
across the specimen during the displacement. The ultimate failure is caused by
predicted compressive damage in the central section of the beam and results in failure
to converge a solution for the analysis.

The simulation result for the 6mm +45/-45 Twintex beam shows a similar response
to the previous analysis, see Figure 6.73. The stiffness is predicted to reduce
progressively due to shear damage up to 130mm where ultimate failure occurs. The
damage contours shown in Figure 6.74 identify fibre direction damage on the
compressive face of the component occurring between 80mm and 120mm
displacement. This corresponds to the ultimate failure predicted by the simulation at
130mm. The shear damage contours also identify a significant amount of shear
damage in the specimen, which unlike the fibre direction damage builds progressively
throughout the analysis.

6-17

The final implicit finite element damage model result is shown in Figure 6.75. The
load displacement response predicted for the 8mm +45/-45 beam, again shows
accurate prediction of stiffness and damage up to a failure at approximately 140mm,
which halts the simulation. The damage contours in Figure 6.76 show that the
ultimate failure in the simulation is caused by fibre direction damage on the
compressive face of the beam, whilst shear damage development is progressive and
spread across the component. This result, like the other +45/-45 beams, showed that
the shear damaging model is accurate for predicting the structural performance of the
beam, but the overall ability of the model is compromised by the inability to converge
a solution after small amounts of fibre direction damage are identified.

6.4.5 Phase 2 Demonstrator Explicit Finite Element Damage Modelling


The explicit finite element damage model presented and calibrated in Chapter 5 was
also used to predictively model all the 1:1 Twintex beams tested. Like the results
presented for the implicit damage model, all the simulations were performed prior to
the testing of the beams and therefore give an accurate representation of the predictive
modelling capability of the technique. The same mesh and boundary conditions, from
the implicit model, were used for the explicit simulations, see Figures 6.63 and 6.64.
The rigid wall was displaced with a constant velocity and nodal damping was applied
to the model to retain stability and reduce dynamic effects during the simulation. A
nodal damping factor of 0.5 was used for all of the Phase 2 demonstrator component
explicit finite element simulations. It should be noted that nodal damping, although
maintaining stability, can affect the results of a simulation and lead to an over stiff
response as well as changing the location and magnitude of damage identified in an
analysis.

Figure 6.77 shows the load displacement response predicted for the 4mm 0/90
Twintex beam compared against the experimental result. The overall performance of
the beam is predicted accurately, with first failure identified at 50mm displacement
and a load of approximately 10kN. The subsequent performance of the beam is also
simulated predictively, with a second major failure occurring just after 200mm
displacement. After this point the model predicts a significant reduction in load
carrying capacity, which is not observed in the test. The other discrepancy between
simulation and experiment is the initial stiffness of the beam, although the analysis
6-18

result shows a slight rise in stiffness after the start of the simulation, which could be
caused by the stiffening effect of the nodal damping applied to the model. This effect
is, as mentioned previously, a potential draw back of using nodal damping as a
method for retaining model stability. The damage contours shown in Figure 6.78
predict the initial failure on the compressive face of the specimen and the second
major failure at a point along the beam, giving a failed specimen with two distinct
areas of critical damage. This compares well to the experimentally observed damage,
shown in Figure 6.49.

Figure 6.79 shows the predicted performance of the 6mm 0/90 Twintex beam
against the experimental test results. Like the implicit damage model, the predicted
stiffness of the beam is significantly higher than that observed during test. The first
major failure is also predicted earlier than was observed, at 50mm rather than 75mm.
The subsequent performance of the beam is also predicted to be stiffer and more
susceptible to damage than the experiment showed. The predicted damage in the
beam, shown in Figure 6.80, suggests that failure occurs initially at the centre of the
beam and subsequently in a second concentrated area along the beam, similar to the
failures observed during test.

The 8mm 0/90 Twintex beam simulation result is shown in Figure 6.81. The
analysis, like that for the 4mm beam, predicts the behaviour of the component fairly
accurately. First major failure is predicted at 50mm and secondary failure at 200mm,
which are both observed experimentally, although only one of the 8mm beams
showed a secondary failure. The initial stiffness of the beams is over-predicted by the
model, but a rise in stiffness in the early stages of the simulation is observed, again
potentially a result of the nodal damping applied. Predicted damage for the 8mm
0/90 Twintex beam is shown in Figure 6.82. Again, the characteristic double
damage zone pattern is observed, although damage is more localised in the central
area when compared to the thinner beams, due to the increased flexural stiffness of
the 8mm beam concentrating the damage zone.

Figure 6.83 shows the result form the first +45/-45 Twintex beam, the 4mm thick
variant. Like the results observed for the Phase 1 demonstrator component, the
damage model initially over-predicts the stiffness of the beam, since damage
6-19

development is not as rapid in the model as the experimental test. This results in a
first failure observed at 50mm for the analysis, which is not seen in the test. After this
point the simulation follows the basic form of the test, but again does not accurately
predict the magnitude of shear damage and its effect on stiffness. The damage
contours shown in Figure 6.84 highlight this, showing localised areas of damage,
similar to those seen for the 0/90 beams. The simulation does show that the section
splays more for the predominantly shear damaging beam, which is an effect observed
during test, see Figure 6.50.

The 6mm +45/-45 Twintex beam, load displacement results are shown in Figure
6.85. The general form of the response is similar to the 4mm beam, although the first
failure is not as marked and the damage development appears more progressive. The
result is close to the experimental result although not identical, due to the nature of the
damage predicted. The damage contour plots for the simulation, shown in Figure 6.86,
highlight the initial central failure predicted in the beam, causing the load reduction at
50mm displacement. Subsequent damage development is over a larger area than
observed in the 4mm beam, although a similar splaying of the section is seen.

The final +45/-45 Twintex beam simulated had an 8mm section thickness. The
result is compared against experiment in Figure 6.87. The simulation curve again
shows an over-stiff response initially, followed by a failure at 50mm. Subsequently
the load rises and a secondary failure is seen at 250mm displacement. The simulation
result again differs from experimental curve due to an over prediction of the beam
stiffness and a subsequent failure behaviour that is not observed in the test. The
damage plots in Figure 6.88 show that the initial failure is caused by an area of
damage in the central section of the beam, which expands during the second stage of
the displacement, between 50mm and 200mm. The splaying in the section for the
8mm beam is predicted to be significantly lower than the 6mm and 4mm beams, due
to the increased section stiffness.

In general, the ability of the explicit damage model to predict damage in the Phase 2
demonstrator component was good for the fibre direction damaging 0/90 Twintex
beams, but was less accurate for the off axis +45/-45 beams.

6-20

6.5

Vehicle Testing

To fully validate the Phase 2 demonstrator component as an alternative to a steel side


intrusion beam, a 6mm thick 0/90 4:1 Twintex beam was installed in the target
vehicle.

The decision to use the 6mm thick 4:1 Twintex beam was made based on test results
for the current steel beam in a large-scale 3-point flexure test, see Figure 6.89. The
beam was selected to give a comparable initial stiffness and failure load. The 4mm
beam was not structurally strong enough and the 6mm 1:1 Twintex beam had underperformed and shown variability in test and was therefore not suitable.

6.5.1 Installation of Beam in Target Vehicle


The beam was installed using the steel brackets detailed in Section 6.4.1. The skin of
the target vehicle door was removed and the brackets were welded to the door
structure. The thermoplastic composite beam was then bolted to the brackets and the
door skin was replaced. A Glass Mat Thermoplastic (GMT) composite door cassette
was mounted on the inside of the door, replacing the steel internal panel. The GMT
door cassette, shown in Figure 6.90, was a design developed previously, during a
manufacturing process research programme, which had not been validated in a vehicle
test [119].

Potentially, a fully thermoplastic door module would be manufactured in a single-shot


moulding process, as a door cassette with fully integrated side intrusion protection.
This test assembly allowed full assessment of the concept without the requirement for
a complex design exercise and the production of expensive tooling.

6.5.2 FMVSS 214 Vehicle Test


The fully assembled door mounted on the target vehicle was tested to the Federal
Motor Vehicle Safety Standard for side impact protection. The test specification
includes a quasi-static intrusion, with prescribed stiffness and maximum load targets,
which was used in this study.

6-21

Both a standard, production model, steel door and side intrusion beam and a door with
a composite beam and door cassette were tested and the results were compared.

The load displacement results from the two tests performed are shown in Figure 6.91.
The initial stiffness, intermediate stiffness and peak load requirements specified in the
test standard are also shown on the figure. The overall performance of the two doors
is very similar and both beams meet the standard requirements. The door with the
composite beam installed shows a drop in load at 75mm displacement, which
corresponds to the first failure crack on the compressive face, observed in the largescale 3-point flexure test. The load continues to rise through to the 305mm maximum
displacement of the test. The standard door with a steel beam shows a slightly
different load displacement response. The initial loading up to 150mm is very linear,
followed by a relatively smooth drop in load, corresponding to the yielding of the
section of the steel beam. The steel beam then re-loads as a tensile strap and yields
significantly again at a displacement of approximately 240mm.

Although the response is different for the two beams, the performance, in terms of the
pass/fail test criteria is very similar. Both the initial and intermediate stiffnesses and
the peak load of the composite beam/door are within 10% of the standard steel door.

6.6

Conclusion

Two demonstrator components have been used during this part of the research, to
further validate the modelling techniques described in Chapter 4 and Chapter 5.

Initially, a first phase component was designed to develop a test methodology and
modelling technique for the beams. The results from an extensive test programme
were compared to simulations using both the ABAQUS and PAM-CRASH models
described earlier in this work. The ABAQUS model was shown to predict the
behaviour of shear damaging components accurately over large displacements,
although for fibre direction damage, the technique was not as applicable. The PAMCRASH model was calibrated predominantly to predict the onset of fibre direction
damage and was therefore more applicable to the 0/90 Twintex beam
configurations.

6-22

For the second phase of the demonstrator component programme, the PAM-CRASH
model was used as a predictive design tool, to develop a new, Phase 2, beam
geometry for a target vehicle. This beam was manufactured and tested using the same
techniques as the Phase 1 beam and results were compared to simulation, which had
been performed predictively, prior to the test work. This allowed the model to be
validated in an industrial type design, manufacture and test exercise. The results
showed that the PAM-CRASH model again performed well when damage was
predominantly in the fibre direction and that the ABAQUS model was less suited to
components where large amounts of damage occur.

To conclude the work, validating the material as a potential alternative to steel for
side intrusion protection and proving the structural integrity of a composite door
concept, a composite beam and door cassette were installed in a target vehicle. This
vehicle was quasi-statically tested to Federal Safety Standard and shown to meet all
the target requirements, with both a steel and composite door.

6-23

6.7

Tables

Beam
Geometry
Phase 1
Phase 1
Phase 1
Phase 1
Table 6.1

0/90
+45/-45
0/90
0/90

Nominal
Thickness
7.5mm
7.5mm
7.5mm
7.5mm

Quantity
Tested
1
1
1
1

Material
1:1 Twintex
1:1 Twintex
4:1 Twintex
3D Twintex
1:1 Twintex
1:1 Twintex
1:1 Twintex
1:1 Twintex
1:1 Twintex
1:1 Twintex
4:1 Twintex
4:1 Twintex
4:1 Twintex
3D Twintex
Steel

Orientation
0/90
+45/-45
0/90
0/90
0/90
0/90
0/90
+45/-45
+45/-45
+45/-45
0/90
0/90
0/90
0/90
-

Nominal
Thickness
7.5mm
7.5mm
7.5mm
7.5mm
8mm
6mm
4mm
8mm
6mm
4mm
8mm
6mm
4mm
6mm
-

Quantity
Tested
2
2
2
2
2
2
2
2
2
2
2
2
2
1
1

Beams tested in large-scale 3-point flexure rig

Beam
Geometry
Phase 2 &
Cassette
Original Beam
Table 6.3

1:1 Twintex
1:1 Twintex
4:1 Twintex
3D Twintex

Orientation

Beams tested in small-scale 3-point flexure rig

Beam
Geometry
Phase 1
Phase 1
Phase 1
Phase 1
Phase 2
Phase 2
Phase 2
Phase 2
Phase 2
Phase 2
Phase 2
Phase 2
Phase 2
Phase 2
Original Beam
Table 6.2

Material

Material
4:1 Twintex
GMT
Steel

Orientation
0/90
Random
-

Beams tested to FMVSS214

6-24

Nominal
Thickness
6mm
-

Quantity
Tested
1
1

6.8

Figures

Figure 6.1

Small-scale 3-point flexure test

Figure 6.2

Large-scale 3-point flexure - Phase 1 beam

6-25

Figure 6.3

Large-scale 3-point flexure - Phase 1 beam

Figure 6.4

Large-scale 3-point flexure - Phase 2 beam

6-26

Figure 6.5

Demonstrator component test vehicle with door cassette fitted

Figure 6.6

FMVSS214 test rig

6-27

Figure 6.7

Typical tubular steel and pressed steel side intrusion beams

Figure 6.8
48mm

Phase 1 demonstrator component geometry, 1050mm x 150mm x

Figure 6.9

Phase 1 demonstrator component central cross section

6-28

6
5

Load (kN)

4
3
2
1
0
0

10

20

30

40

50

60

70

Displacement (mm)

Figure 6.10
test result

1:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure

Splaying (mm)

3
2

1
0
0

10

20

30

40

50

60

70

Displacement (mm)

Figure 6.11 1:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure
test section splaying

6-29

Load (kN)

0
0

10

20

30

40

50

60

70

80

90

Displacement (mm)

Figure 6.12 1:1 Twintex +45/-45 Phase 1 demonstrator small-scale 3-point


flexure test result

Splaying (mm)

3
2

1
0
0

10

20

30

40

50

60

70

Displacement (mm)

Figure 6.13 1:1 Twintex +45/-45 Phase 1 demonstrator small-scale 3-point


flexure test section splaying

6-30

6
5

Load (kN)

4
3
2
1
0
0

10

20

30

40

50

60

Displacement (mm)

Figure 6.14
test result

4:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure

Splaying (mm)

4
3
2
1
0
0

10

20

30

40

50

60

70

Displacement (mm)

Figure 6.15 4:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure
test section splaying

6-31

7
6

Load (kN)

5
4
3
2
1
0
0

10

20

30

40

50

60

70

80

Displacement (mm)

Figure 6.16
test result

3D Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure

Splaying (mm)

3
2

1
0
0

10

20

30

40

50

60

70

Displacement (mm)

Figure 6.17 3D Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure


test section splaying

6-32

Load (kN)

20
18
16
14
12
10
8
6
4
2
0

0/90 Beam 1
0/90 Beam 2

50

100

150

200

250

300

Displacement (mm)

Load (kN)

Figure 6.18
test results

1:1 Twintex 0/90 Phase 1 demonstrator large-scale 3-point flexure

20
18
16
14
12
10
8
6
4
2
0

45/45 Beam 1
45/45 Beam 2

50

100

150

200

250

300

Displacement (mm)
Figure 6.19 1:1 Twintex +45/-45 Phase 1 demonstrator large-scale 3-point
flexure test results

6-33

Load (kN)

20
18
16
14
12
10
8
6
4
2
0

4:1 Beam 1
4:1 Beam 2

50

100

150

200

250

300

Displacement (mm)

Load (kN)

Figure 6.20
test results

4:1 Twintex 0/90 Phase 1 demonstrator large-scale 3-point flexure

24
22
20
18
16
14
12
10
8
6
4
2
0

3D Beam 1
3D Beam 2

50

100

150

200

250

300

Displacement (mm)

Figure 6.21
test results

3D Twintex 0/90 Phase 1 demonstrator large-scale 3-point flexure

6-34

Figure 6.22

Phase 1 demonstrator small-scale 3-point flexure model

Figure 6.23

Phase 1 demonstrator large-scale 3-point flexure model

6-35

6
Experimental Test

Load (kN)

ABAQUS Model

4
3
2
1
0
0

10

20

30

40

50

60

70

Displacement (mm)

Figure 6.24 1:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure
test implicit finite element simulation
FV1 & 3 - 50mm, 60mm, 70mm

Figure 6.25 1:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure
test implicit finite element simulation - predicted damage

6-36

4
Experimental Test
ABAQUS Model

Load (kN)

0
0

10

20

30

40

50

60

70

80

90

100

Displacement (mm)

Figure 6.26 1:1 Twintex +45/-45 Phase 1 demonstrator small-scale 3-point


flexure test implicit finite element simulation
FV 3 - 20mm, 40mm, 60mm, 80mm, 100mm

Figure 6.27 1:1 Twintex +45/-45 Phase 1 demonstrator small-scale 3-point


flexure test implicit finite element simulation - predicted damage

6-37

Load (kN)

24
22
20
18
16
14
12
10
8
6
4
2
0

0/90 Beam 1
0/90 Beam 2
ABAQUS Model

50

100

150
200
Displacement (mm)

250

300

Figure 6.28 1:1 Twintex 0/90 Phase 1 demonstrator large-scale 3-point flexure
test implicit finite element simulation
FV1 & 3 - 10mm, 20mm, 30mm, 40mm, 50mm

Figure 6.29 1:1 Twintex 0/90 Phase 1 demonstrator large-scale 3-point flexure
test implicit finite element simulation - predicted damage

6-38

Load (kN)

24
22
20
18
16
14
12
10
8
6
4
2
0

45/45 Beam 1
45/45 Beam 2
ABAQUS Model

50

100

150

200

250

300

Displacement (mm)

Figure 6.30 1:1 Twintex +45/-45 Phase 1 demonstrator large-scale 3-point


flexure test implicit finite element simulation
FV1 & 3 - 140mm, 200mm, 240mm, 280mm

Figure 6.31 1:1 Twintex +45/-45 Phase 1 demonstrator large-scale 3-point


flexure test implicit finite element simulation - predicted damage

6-39

Experimental Test

PAM Crash Model


Load (kN)

5
4
3
2
1
0
0

10

20

30

40

50

60

70

80

90

100

Displacement (mm)

Figure 6.32 1:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure
test explicit finite element simulation
30mm, 50mm, 70mm, 90mm

Figure 6.33 1:1 Twintex 0/90 Phase 1 demonstrator small-scale 3-point flexure
test explicit finite element simulation - predicted damage

6-40

4
Experimental Test
PAM Crash Model

Load (kN)

0
0

10

20

30

40

50

60

70

80

90

100

Displacement (mm)

Figure 6.34 1:1 Twintex +45/-45 Phase 1 demonstrator small-scale 3-point


flexure test explicit finite element simulation
30mm, 50mm, 70mm, 90mm

Figure 6.35 1:1 Twintex +45/-45 Phase 1 demonstrator small-scale 3-point


flexure test explicit finite element simulation - predicted damage

6-41

24
22

0/90 Beam 1

20
18

0/90 Beam 2
PAM Crash Model

Load (kN)

16
14
12
10
8
6
4
2
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.36 1:1 Twintex 0/90 Phase 1 demonstrator large-scale 3-point flexure
test explicit finite element simulation
40mm, 120mm, 200mm, 280mm

Figure 6.37 1:1 Twintex 0/90 Phase 1 demonstrator large-scale 3-point flexure
test explicit finite element simulation - predicted damage

6-42

Load (kN)

24
22
20
18
16
14
12
10
8
6
4
2
0

45/45 Beam 1
45/45 Beam 2
PAM Crash Model

50

100

150

200

250

300

Displacement (mm)

Figure 6.38 1:1 Twintex +45/-45 Phase 1 demonstrator large-scale 3-point


flexure test explicit finite element simulation
40mm, 120mm, 200mm, 280mm

Figure 6.39 1:1 Twintex +45/-45 Phase 1 demonstrator large-scale 3-point


flexure test explicit finite element simulation - predicted damage

6-43

Concept

Thickness
(mm)

8.0

8.0

8.0

8.0

Mass (kg)

3.3

3.5

3.9

4.0

1st Failure
Load (kN)

12.5

19.5

13.0

24.0

1st Failure
Disp. (mm)

78

83

60

52

Geometry

Figure 6.40 Summary of concept development analysis results for Phase 2


demonstrator component

Figure 6.41
55mm

Phase 2 demonstrator component geometry, 1020mm x 268mm x

6-44

Figure 6.42

Phase 2 demonstrator component central cross section

Figure 6.43
model

Phase 2 demonstrator component installation in target vehicle CAD

Figure 6.44
installation

Phase 2 demonstrator component profile modification for vehicle

6-45

Figure 6.45

Phase 2 demonstrator steel brackets for vehicle installation

Figure 6.46
thickness

Phase 2 demonstrator tool showing insert to reduce component

Figure 6.47
thickness

Phase 2 demonstrator tool showing insert to reduce component

6-46

1.

Preconsolidated
blanks heated to
200C in oven.

2.

Material
transferred to
cool tool.

3.

Pressure applied
for 90s as
material cools.

4.

Tool opened and


formed part
removed.

Figure 6.48

Moulding process for Phase 2 demonstrator component

6-47

Figure 6.49 1:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure tested specimen

Figure 6.50 1:1 Twintex +45/-45 4mm Phase 2 demonstrator large-scale 3-point
flexure tested specimen

Figure 6.51 4:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure tested specimen

Figure 6.52 3D Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point


flexure tested specimen

6-48

40
0/90 4mm Beam 1

35

0/90 4mm Beam 2

Load (kN)

30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.53 1:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure test results

35
0/90 6mm Beam 1

30

0/90 6mm Beam 2

Load (kN)

25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.54 1:1 Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point
flexure test results

6-49

Load (kN)

45
40

0/90 8mm Beam 1

35

0/90 8mm Beam 2

30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.55 1:1 Twintex 0/90 8mm Phase 2 demonstrator large-scale 3-point
flexure test results

35
45/45 4mm Beam 1

30

45/45 4mm Beam 2

Load (kN)

25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.56 1:1 Twintex +45/-45 4mm Phase 2 demonstrator large-scale 3-point
flexure test results

6-50

50
45

45/45 6mm Beam 1

40

45/45 6mm Beam 2

Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.57 1:1 Twintex +45/-45 6mm Phase 2 demonstrator large-scale 3-point
flexure test results

50
45

45/45 8mm Beam 1

40

45/45 8mm Beam 2

Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.58 1:1 Twintex +45/-45 8mm Phase 2 demonstrator large-scale 3-point
flexure test results

6-51

45
4:1 4mm Beam 1

40

4:1 4mm Beam 2

Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.59 4:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure test results

50
45

4:1 6mm Beam 1

40

4:1 6mm Beam 2

Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.60 4:1 Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point
flexure test results

6-52

Load (kN)

55
50
45
40
35
30
25
20
15
10
5
0

4:1 8mm Beam 1


4:1 8mm Beam 2

50

100

150

200

250

300

Displacement (mm)

Load (kN)

Figure 6.61 4:1 Twintex 0/90 8mm Phase 2 demonstrator large-scale 3-point
flexure test results

55
50
45
40
35
30
25
20
15
10
5
0

3D 6mm Beam 1
3D 6mm Beam 2

50

100

150

200

250

300

Displacement (mm)

Figure 6.62 3D Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point flexure
test results

6-53

Figure 6.63

Phase 2 demonstrator large-scale 3-point flexure model

Figure 6.64

Phase 2 demonstrator large-scale 3-point flexure model

6-54

40
0/90 4mm Beam 1
0/90 4mm Beam 2
ABAQUS Model

35

Load (kN)

30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.65 1:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation

FV1 & 3 - 20mm, 30mm, 40mm

Figure 6.66 1:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation - predicted damage

6-55

35
0/90 6mm Beam 1
0/90 6mm beam 2
ABAQUS Model

30

Load (kN)

25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.67 1:1 Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation

FV1 & 3 - 20mm, 30mm, 40mm

Figure 6.68 1:1 Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation - predicted damage

6-56

45
0/90 8mm Beam 1
0/90 8mm Beam 2
ABAQUS Model

40

Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.69 1:1 Twintex 0/90 8mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation

FV1 & 3 - 20mm, 30mm, 35mm

Figure 6.70 1:1 Twintex 0/90 8mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation - predicted damage

6-57

35
45/45 4mm Beam 1
45/45 4mm beam 2
ABAQUS Model

30

Load (kN)

25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.71 1:1 Twintex +45/-45 4mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation

FV1 & 3 - 40mm, 80mm, 120mm

Figure 6.72 1:1 Twintex +45/-45 4mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation - predicted damage

6-58

Load (kN)

50
45

45/45 6mm Beam 1

40

45/45 6mm Beam 2

35

ABAQUS Model

30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.73 1:1 Twintex +45/-45 6mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation

FV1 & 3 - 40mm, 80mm, 120mm

Figure 6.74 1:1 Twintex +45/-45 6mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation - predicted damage

6-59

50
45/45 8mm Beam 1
45/45 8mm Beam 2
ABAQUS Model

45
40
Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.75 1:1 Twintex +45/-45 8mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation

FV1 & 3 - 40mm, 80mm, 120mm, 140mm

Figure 6.76 1:1 Twintex +45/-45 8mm Phase 2 demonstrator large-scale 3-point
flexure implicit finite element simulation - predicted damage

6-60

40
0/90 4mm Beam 1
0/90 4mm Beam 2
PAM Crash Model

35

Load (kN)

30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.77 1:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation

40mm, 120mm, 200mm, 280mm

Figure 6.78 1:1 Twintex 0/90 4mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation - predicted damage

6-61

35
0/90 6mm Beam 1
0/90 6mm beam 2
PAM Crash Model

30

Load (kN)

25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.79 1:1 Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation

40mm, 120mm, 200mm, 280mm

Figure 6.80 1:1 Twintex 0/90 6mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation - predicted damage

6-62

45
0/90 8mm Beam 1
0/90 8mm Beam 2
PAM Crash Model

40

Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.81 1:1 Twintex 0/90 8mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation

40mm, 120mm, 200mm, 280mm

Figure 6.82 1:1 Twintex 0/90 8mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation - predicted damage

6-63

35
45/45 4mm Beam 1

30

45/45 4mm beam 2


PAM Crash Model

Load (kN)

25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.83 1:1 Twintex +45/-45 4mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation

40mm, 120mm, 200mm, 280mm

Figure 6.84 1:1 Twintex +45/-45 4mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation - predicted damage

6-64

50
45

45/45 6mm Beam 1


45/45 6mm Beam 2
PAM Crash Model

40
Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.85 1:1 Twintex +45/-45 6mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation

40mm, 120mm, 200mm, 280mm

Figure 6.86 1:1 Twintex +45/-45 6mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation - predicted damage

6-65

50
45/45 8mm Beam 1
45/45 8mm Beam 2
PAM Crash Model

45
40
Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.87 1:1 Twintex +45/-45 8mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation

40mm, 120mm, 200mm, 280mm

Figure 6.88 1:1 Twintex +45/-45 8mm Phase 2 demonstrator large-scale 3-point
flexure explicit finite element simulation - predicted damage

6-66

20
18

Steel Beam

16
Load (kN)

14
12
10
8
6
4
2
0
0

50

100

150

200

250

300

Displacement (mm)

Figure 6.89

Steel side intrusion beam large-scale 3-point flexure test

Figure 6.90
test

GMT door cassette component used in FMVSS214 composite beam

6-67

50
45

Steel Beam

40

Composite Beam

Load (kN)

35
30
25
20
15
10
5
0
0

50

100

150

200

250

Displacement (mm)

Figure 6.91

FMVSS214 vehicle side intrusion test - load displacement results

6-68

300

Chapter 7
7.1

Discussion and Conclusions

Introduction

Thermoplastic composite materials are becoming a viable alternative to steel and


aluminium for use in semi-structural applications in the automotive industry.
Limitations in the current understanding of their damaging behaviour and a lack of
fully validated modelling tools, is though, a barrier to their application in fully
crashworthy components. The aim of this work was to develop a predictive damage
modelling capability for a thermoplastic composite material, namely Twintex and to
use the results for the design and test of a structural crashworthy component.

Two approaches to damage modelling have been investigated, the first using an
implicit finite element code and the second using an explicit finite element code. Both
have been calibrated with a combination of published and experimental test data and
validated for a range of in-plane damage scenarios.

In general, testing numerous specimens for calibration of complex material models is


costly in terms of both time and expense. It is therefore important, for the purposes of
industrial acceptance, that it is as simple as possible to gather the necessary data to
facilitate the accurate calibration of damage models. It is also imperative, that where
applicable, material data that is already available can be used for these purposes. It is
for this reason, that the current body of work has not been an exercise in the
development of calibration regimes or a thorough characterisation of the candidate
material. In fact, the author has where possible, used currently available data and
where not, used what was judged to be the least complex test standard or method to
gather the necessary data.

The validated models have then been applied to the design of an industrial
demonstrator component to investigate the potential applications and limitations of
the techniques. The demonstrator component has finally been tested to prove that
thermoplastic composite materials, if designed correctly, can offer an alternative to
high strength steel for crashworthy applications.

7-1

In this section, the calibration test methods, damage models used and component
testing undertaken are discussed. Limitations with the current work are investigated
and suggestions are made for further work to improve the techniques developed.

7.2

Coupon Test Methods for Calibration and Validation of Finite

Element Damage Models


Since Twintex, is a 0/90 balanced weave fabric reinforced composite, there are only
three basic tests required to calibrate the critical parameters for the in-plane behaviour
of the material. These are, a 0 tensile tests, a 0 compression test and a shear test. In
the current work, a +45/-45 tensile test was selected for the shear characterisation,
as both specimen preparation and test rig development were significantly simpler.
This method has been validated by previous authors [19] and has also been published
as an ISO standard method [120]. Developments of these tests, using specimens with
stress concentrations were used to further validate numerical models after calibration.

7.2.1 Quantity of Specimens Tested and Scatter of Results


Material availability for preparation of test specimens was an important factor
throughout the research, up until the demonstrator component programme, when
increased stocks became available. In general for each test undertaken, at least five
specimens were prepared to allow for variation in material performance. Ideally, more
specimens of each type would have been tested, although this was not possible.

For the tensile 0/90 test, results from 11 successful tests from 15 specimens are
presented. Since the same method was used for all subsequent tensile tests, both with
and without stress concentrations, it is hypothesised that the variability observed
during the 0/90 tensile test would be similar to the levels observed in subsequent
tests. In this first set of tests, ultimate strength ranged form 244 MPa to 302 MPa,
which equates to approximately 10% about the mean value of 279 MPa. This
suggests that the variable nature of the manual transfer, non-isothermal processing
method could introduce a certain amount of scatter in observed test results. In
consequence, the exact prediction of the performance of a component, using
numerical methods will always be dependant on, not only the quality of the simulation

7-2

techniques, but also on the quality of the composite material, which has in turn been
shown to be highly dependant on manufacturing process [11][12].

7.2.2 Strain Measurement Techniques


In terms of experimental technique, the method for measurement of strain during test
was where the present authors approach differed from previous studies [15][16]. In all
of the tensile tests undertaken, strain measurement was performed using an
extensometer mounted on the specimen, measuring nominal strain over a 50mm
gauge length. This was essential for two reasons. Firstly the quality of bond
achievable between a strain gauge and a polypropylene matrix composite specimen is
highly variable, depending on the surface treatment and preparation and secondly,
even with excellent bond quality, the gauge will only measure over a short strain
range before becoming detached. When investigating a thermoset matrix composite,
where behaviour is often elastic with a brittle failure, strain gauges are acceptable for
material characterisation. In the case of polypropylene Twintex, especially in sheardominated tests, where behaviour was highly non-linear and the extension over a
50mm gauge length was observed to be greater than 15%, strain gauges are less
applicable.

The only characterisation test to use a strain gauge rather than an extensometer, was
the 0/90 compressive test. This test captured the linear portion of the material
behaviour accurately, but failed to give strain behaviour after the first critical failure,
when the strain gauge became detached. This resulted in a phase of specimen
compressive damage not being captured by the test result.

7.2.3 Shear Behaviour During Off Axis Tests


The shear damaging behaviour of Twintex, was difficult to model. In the early stages
of the work, testing of tensile and compressive specimens manufactured from +45/45 mouldings, showed similar results, in terms of the modulus and shear damaging
behaviour. This allowed modelling, especially using the implicit technique described
in Chapter 4, to be calibrated to reflect observed material performance.

When comparing these results to the behaviour of the +45/-45 specimens with stress
concentrations, an interesting phenomenon is observed. The compressive off-axis test
7-3

with a stress concentration appears to behave in a similar way to the plain


compressive specimen, with some initial shear deformation followed by a
combination of shear deformation and out of plane buckling. The tensile +45/-45
specimen with a stress concentration, on the other hand, provides results which are
more difficult to predict. In these specimens, a stiffer than expected response is
observed, with failure occurring at significantly lower strain than in the specimen
without a stress concentration. This result is due to the method of strain measurement,
over a 50mm gauge length. When the gauge is measuring a strain of 5%, strain levels
could be locally higher than 25% if all the deformation is concentrated around the
hole.

7.2.4 Out of Plane Deformation During Compressive Test


Due to the nature of the polypropylene matrix material, it was particularly difficult to
characterise the damaging behaviour during compressive tests. The damage models
are predominantly concerned with single, multi-layered, shell element representations
of the composite and as such do not accurately model the complex out of plane and
buckling damage seen during test. This, for thermoset matrix composites is not a
particularly critical issue, since the material is loaded and deforms linearly until
delamination, fibre buckling or catastrophic matrix failure occur and the specimen
fails. With a polypropylene matrix the specimen starts to deform out of plane, without
a catastrophic drop in load carrying capability. For this reason it was difficult in some
cases to differentiate the contribution of in-plane and out of plane damage
development during these tests.

7.2.5 Damage Development in Hole in Plate Specimens


The hole in plate specimen as a method for investigation of damage in a non-uniform
geometry and as a validation for numerical studies has been shown to be a valuable
tool. The specimen itself is little more complex than a standard tensile or compressive
specimen and requires virtually no modification in method for testing.

If it is accepted that the step from standard test specimen results for material
characterisation to prediction of composite component behaviour is too great for
current simulation techniques, the hole in plate type specimen could be an
economically viable midway point. This is especially true when considering the
7-4

results of the later stages of this work, which show that for thermoplastic composite
materials, an accurate in-plane model for the fibre direction behaviour can yield
encouraging results in terms of the prediction of damage development in complex
geometries.

7.3

Implicit Finite Element Damage Modelling

The purpose of this section of the work was to consider the potential of a previously
proposed damage model for thermoset matrix composites, for the simulation of
accumulated damage in thermoplastic composites. During the initial stages of the
investigation it was quickly identified that the models treatment of the brittle matrix
material and shear failure criteria yielded poor results when applied to thermoplastics.
This led to the development of a simplified model for thermoplastic matrix
composites, using a ply based maximum strength criteria for the fibre direction and a
shear degradation regime and ultimate failure strain. The only component of the
model retained from the work of Chang [68][79] was the shear modulus degradation
relationship.

Subsequent work was then undertaken to validate the new model and investigate the
potential of such an approach to be used as a composite material design tool. It should
be noted that the purpose of this work was not to provide an alternative to explicit
dynamic modelling techniques, but to investigate the potential for an implicit code to
be used to provide a complementary solution and a numerically less expensive design
capability.

7.3.1 The Treatment of Fibre Direction and Shear Damage


The most interesting feature of this model when comparing to the other model
investigated during this research programme is the separate treatment of fibre
direction and shear damage. When considering a brittle matrix composite, where
failure strains in both fibre direction and shear are within similar ranges, it is possible
to consider a damage state for a unit volume and degrade the elastic constants
accordingly. In a woven fibre reinforced thermoplastic matrix composite, this can
potentially lead to situations where shear damage, which in reality would result in
fairly minimal reduction in fibre direction properties, results in levels of damage that

7-5

significantly reduce the load carrying capacity of the ply. The separation of the two
avoids this situation and in the case of coupon test simulations, offered promising
results.

7.3.2 Accuracy of Fibre Direction Damaging Behaviour


Using the implicit damage model it was possible to gain excellent agreement between
tensile experiment and test in the fibre direction. It was also possible to accurately
model the compressive behaviour of Twintex, where simulation again matches closely
experimentally observed behaviour. The implicit model does however fail
catastrophically at the point when stress levels reach the calibrated compressive
failure level. This appears to match the experimental result well, although it should be
noted that, as mentioned previously, the specimen in reality continued to deform out
of plane after this first ply failure, which resulted in a significant load drop, but not in
catastrophic collapse.

This result, for the compressive damage prediction, becomes more significant when
considering more complex geometries, where ultimate failure in a localised region
would not necessarily lead to global collapse of a structure.

For the fibre direction specimen with a stress concentration, ultimate failure is also
predicted prematurely, at a stress level approximately 15% below that observed in
test. This suggests that inclusion of a residual strength in the model could improve the
stability of the solution.

7.3.3 Accuracy of Shear Damaging Behaviour


The accuracy of the shear damaging behaviour of the model, was good for the +45/45 tensile test specimen. This was expected, since this test was used to calibrate the
shear damage parameter . In the off-axis compression test, the result of simulation
was not as encouraging. As mentioned previously, the specimen deformed both in and
out of plane during this test, resulting in a damaging behaviour that was not wholly
captured by the model. This result was also seen in the +45/-45 compression
specimen with a hole, where, especially at high levels of displacement, the model
predicts a stiffer response. It is interesting to note, that the 5mm displacement

7-6

observed at the completion of this test, equates to a global strain level of


approximately 4 %, when considering the whole specimen. Similar work presented by
Chang and Lessard [79] shows results up to between 0.6% and 1.6% strain. If these
levels were considered for the work presented here, excellent agreement would be
seen. It is therefore concluded that the model is accurate for in-plane damaging
modes, although buckling deformation, as expected, cannot be accounted for
accurately.

The result for the tensile +45/-45 specimen with a hole is less encouraging. The
model significantly under predicts the response of specimen in terms of stiffness and
ultimate failure load. This is due in part to the model predicting shear damage
throughout the specimen, which to a certain extent was not observed in test, where the
deformation was more localised around the hole. It is also possible that fibre reorientation around the hole, which is not accounted for accurately by the model, could
influence the performance at high strain levels. In this case as the fibres shear in the
area of the hole, they become oriented in the direction of the test. Due to the ductility
of the matrix, the composite can still function as a homogeneous material and hence
the local stiffness can remain high.

7.3.4 Application of the Implicit Damage Model to a Complex Component


During the development of the automotive demonstrator components manufactured
and tested during this research programme, the model was used to predictively
simulate the damage behaviour, with varying degrees of success. For the Phase 1
demonstrator geometry, the model accurately predicted the small-scale 3-point flexure
behaviour of both the 0/90 and the +45/-45 lay-up beams. These results were for
relatively small displacement, around 80mm, compared to later tests up to 300mm.

It is in these later tests that limitations of the technique become apparent, for the
0/90 beams in particular, where the model predicts catastrophic failure at a
displacement of 50mm. In reality, although damage had occurred, the beams
continued to carry load until an ultimate failure displacement of up to 300mm. This
limitation of the model is in part due to the inability of an implicit code to converge a
solution when elements have deformed significantly due to reduction in load carrying

7-7

capacity and also due to the nature of the damage model, where compressive failure
leads to a reduction of the modulus to a nominal value.

In the case of the +45/-45 Phase 1 beam simulations, the results are more
encouraging. Although the model under predicts the stiffness of the beam towards the
end of the test, the solution remains stable and converges up to a displacement of
300mm.

For the Phase 2 beam geometry, similar results are observed. The 0/90 beam
simulations all predict failure to occur catastrophically during the early stages of the
test, a phenomenon that was not observed experimentally. It is also interesting to note
that for the 6mm beam, the stiffness, even during the initial stages of test is not
correctly predicted. It is felt that this was due to moulding issues associated with the
6mm beam, which will be discussed later, and not due to the modelling technique.

The +45/-45 Phase 2 beam simulations, like the Phase 1 results, show good
correlation with test, although the simulations all predict failure at incorrect
displacements. In all cases, fibre direction damage predicted in the beams is resulting
in failure to converge a solution at approximately 120mm displacement.

7.3.5 Suitability

of

the

Implicit

Finite

Element

Method

for

Large

Displacements
In general, attempting to capture and model highly non-linear material behaviour
using an implicit finite element code could be seen as a futile exercise. The nature of
the technique is such that if failure is predicted and large displacement is expected,
the step size of each iteration, which has to be small enough to allow the solution to
converge, could potentially result in a numerically uneconomical solution.

Results from this work have shown that this assumption does not necessarily hold true
for all situations. The progressive nature of the shear damage model allows the
technique to work effectively in certain cases, producing stable results for the +45/45 Phase 1 side intrusion beam simulation. For components where damage is
predominantly in the fibre direction the model is less effective. Although it is based
on observed phenomena, the failure model causes instability, by reducing elastic
7-8

constants to levels where a solution cannot be converged satisfactorily in the specified


step.

7.4

Explicit Finite Element Damage Modelling

In the second part of the numerical modelling research work presented, the degenerate
bi-phase damage model implemented in a commercial finite element analysis code
was investigated. This model, as discussed in Chapter 2, has been used with a certain
degree of success to model thermoset matrix composites with various fibre
reinforcements, but little work has been published on its use to predict the behaviour
of thermoplastic matrix composites.

The aim of the work was to calibrate and validate a model for polypropylene Twintex,
with the in-plane damaging behaviours observed in simple tests and to then use this
calibrated model for the analysis of more complex geometries.

7.4.1 The Treatment of Fibre Direction and Shear Damage


Unlike the previous numerical modelling work, the calibration of this model was
limited to the parameters and material law available in the PAM-CRASH code. For
this reason, the model was calibrated to offer a general representation of the material
behaviour and does not appear to show as good an agreement with experiment for the
coupon tests as the implicit damage model developed by the author.

This however, it is shown, does not necessarily compromise the ability of the model
when considering more complex geometries.

7.4.2 Accuracy of Fibre Direction Damaging Behaviour


The nature of the bi-linear damage law and relation between damage and elastic
constant degradation lead to a characteristic, curved response, from a simulation of
the tensile test for a 0/90 specimen. Although this is not an exact representation of
the linear-elastic and then catastrophically damaging behaviour observed during test,
the energy to failure is similar for both curves. During the calibration of the model, it
was not possible to achieve a better fit for this simulation.

7-9

For the 0/90 compressive test simulation, the linear behaviour up to the point of
major failure is captured accurately. It is however noted that in the simulation, the
specimen is predicted to carry load up until failure at 3% strain. It is difficult to
confirm this behaviour against test, since the stress/strain data available is up till the
point at which the strain gauge became detached and not ultimate failure in the
specimen.

The tensile 0/90 specimen with a hole stress concentration also shows an interesting
result. Here, at approximately 1% strain, when the first damage is identified in the
specimen, catastrophic failure occurs rapidly. This is due to load redistribution across
the central section of the specimen as damage accumulates around the hole. The
nature of the damage model, with an initial damage point marking the start of a
progressive failure behaviour, is such that this type of simulation, for a specimen with
a hole will always yield a similar result, with catastrophic failure occurring soon after
damage initiation. In more complex geometries, where load redistribution is possible
without catastrophic consequences, this type of behaviour is not of great concern.

7.4.3 Accuracy of Shear Damaging Behaviour


The shear damaging response of the material model was more difficult to calibrate
accurately, since any variation in the damage parameters to obtain an accurate shear
damage behaviour, had repercussions in terms of the response of the fibre direction
damage model. It was also very difficult, regardless of the parameters selected, to
calibrate the model to remain stable at higher strain levels. In fact after an extensive
calibration simulation programme the final set of parameters selected were chosen
primarily based on the fibre direction damage.

This difficulty in calibrating the degenerate bi-phase model to accurately capture


shear damaging behaviour, especially for woven fabric reinforced composites, has
been discussed by McCarthy and Wiggenraad [113]. The phenomenon which has
previously been observed for thermoset matrix composites was expected and seen to
be more pronounced for thermoplastic composites where the matrix is tougher and
displays a more ductile behaviour during deformation.

7-10

Since the behaviour of the +45/-45 tensile test simulation was compromised due to
the calibration of the model, it was expected that a similar result would be observed
for the off-axis compression test. This was the case, with the same, over stiff
response, and premature ultimate failure.

The combination of this inaccuracy in the modelling of shear damage and the
instability of the test simulations with a stress concentration, yielded poor agreement
with the results for the off axis tensile and compressive test with a stress
concentration. Again, this phenomenon was both expected and unavoidable due to the
nature of the model and the tests.

7.4.4 Sensitivity to Calibration Parameters


During the calibration phase of the model, as previously discussed, it was the
objective of the work to develop and propose a calibration scheme, which accurately
captured the full range of in-plane damage modes for Twintex. During the study, it
became apparent that this would not be fully achievable. Eventually a calibration
scheme was developed to primarily reflect the fibre direction damage of the
composite, whilst remaining stable during shear damage, since it was felt that in a
structural component, the fibre direction damage behaviour would be dominant.

The sensitivity analysis presented shows the effect of variation in the various
parameters, for the simulation of the 0/90 and +45/-45 tensile test. It was shown in
this part of the work that there was little possibility for the model to accurately
capture shear behaviour without significantly compromising the fibre direction
behaviour, as the 0/90 simulation result is clearly more sensitive to the variation of
damage parameters.

7.4.5 Application of the Explicit Damage Model to a Complex Component


Although the model had been shown to be truly accurate only for coupon simulations
with predominantly fibre direction damage, the calibrated model was used to simulate
all the 1:1 balanced twill weave demonstrator components tested. This included
beams with 0/90 and with +45/-45 fibre architectures.

7-11

The results from the Phase 1 demonstrator component showed excellent agreement
for the 0/90 beams, both in the small-scale and large-scale 3-point flexure tests. In
the large-scale test simulations, the prediction of the first critical damaging event and
subsequent load drop and reloading behaviour closely matched the experimental
result. The predicted damage zones also match closely, suggesting that the model is
highly applicable to damaging situations of this type. The +45/-45 beam
simulations, as expected, do not show as good agreement with experimental test. In
this case, the predicted behaviour is over stiff with subsequent damage resulting in a
drop in load carrying capacity. This is a similar effect to that seen in the coupon test
simulations and is due to the model not capturing the shear damaging behaviour of the
composite correctly. It was therefore concluded from this part of the demonstrator
component programme that the predictive simulation of shear damage development in
a component was not accurate using the current calibration of the model.

For the Phase 2 demonstrator component, where the model was being used fully
predictively, the simulation results for the 0/90 beams, again showed good
agreement with test. The 6mm beam analysis did however highlight that the test had
provided results which were uncharacteristically poor in terms of bending stiffness.
The +45/-45 beam simulations in this phase of the demonstrator work, again
showed poor correlation with test, as expected.

In general, the model, which was calibrated primarily for fibre direction damage
behaviour, performed well. In the second phase of the demonstrator component
programme, the model was used fully predictively and yielded acceptable results.

7.4.6 Applicability of Damage Model to Thermoplastics


The degenerate bi-phase model has been shown to be applicable to thermoplastic
matrix composites, but with certain limitations, which have been highlighted by this
work. The damage model is not particularly suitable for woven fabric reinforced
composites, especially with ductile matrix materials, which was highlighted during
the calibration, validation and demonstrator component simulations.

In summary, during the calibration of the damage parameters it had been shown that
capturing shear and fibre direction damage with a single range of parameters was not
7-12

possible for Twintex, due to the woven fabric reinforcement and the ductile matrix.
This was confirmed by the demonstrator component simulations. The discrepancy
between test and analysis for the +45/-45 beams was however not as great as
initially expected. This suggests that if the limitations of the technique are considered
and it is used in situations where damage is predominantly in the fibre direction, there
is potential to use the model as a predictive design tool.

7.4.7 Suitability of the Explicit Finite Element Method for Large Displacements
Although the damage model was implemented in a commercial code and the user had
no specific control over the numerical treatment of the damage development in the
composite, the method yielded interesting and in certain cases, successful results. In
particular, when considering a complex component, the explicit finite element
technique was clearly far more suited to dealing with damage and material nonlinearity than the previous implicit finite element damage modelling work.

7.5

Application of Thermoplastic Composites to Crashworthy

Automotive Structures
Polypropylene matrix based composite materials have previously been used
successfully in semi-structural automotive applications. The demonstrator component
programme undertaken during this study has shown that they also have the potential
to be used in fully crashworthy applications. A glass reinforced polypropylene
composite side intrusion beam has been shown to offer a similar level of performance
to that of a current steel beam design. There is a potential weight penalty if the
composite beam is used as a direct replacement for a current steel component, but if
the beam is considered as part of a structural thermoplastic composite door module it
may become a viable alternative. Typically, the masses of the Phase 2 demonstrator
beams tested ranged from 2.1kg to 4.3kg, with the final vehicle test performed on an
8mm thick, 4.2kg beam. This is a significant weight penalty when directly comparing
the composite beam to the 1.9kg steel beam from the target vehicle, even when
considering that the geometry and lay-up of the composite design had not been fully
optimised.

7-13

7.5.1 Industrial Processing Techniques for Thermoplastic Composite


Various authors have presented idealised processing parameters for thermoplastic
composite materials, to give optimum material properties and produce mouldings of
the highest quality. This can be achieved through careful control of preheat time and
temperature, transfer time, moulding pressure, tool temperature and consolidation
time. These idealised parameters may be more difficult to achieve in an industrial
process. During most of the moulding that was undertaken for this work, a semiautomated procedure was used, where pre-heat was controlled using an industrial
oven, measuring material surface temperature and the press closure cycle was
automated.

Slight over prediction of strength and stiffness, during the modelling of the
demonstrator components could be explained by these parts having a marginally
poorer moulding quality than the laboratory manufactured test specimens, used for
material model calibration. This is reinforced when considering the quality of
moulded specimens using optical microscopy. Typically the void content observed in
the 1:1 Twintex beams, manufactured for the Phase 1 demonstrator component
programme, ranged from 1.8% to 6.8% across the section of the beam [121], which
compares to 2% observed in 1:1 Twintex plaques manufactured using a laboratory
scale process [27]. It is also noted that the 3D Twintex beams, which consistently
showed higher performance when compared to 1:1 Twintex beams, also had the
lowest void content, with a range of between 0.5% and 4.5% across a typical beam
[121]. It is likely that this is due to the two stage moulding technique applied to the
3D Twintex, where an extra pre-consolidation of the fabric was included prior to final
moulding. This was not required for either the 1:1 or 4:1 Twintex beams, as the
material blanks were supplied pre-consolidated.

The development of tooling including a combined sprung blankholder and shear edge
was novel in terms of industrial application. This technical innovation allows variable
thickness co-moulding of fabrics and flowing thermoplastic composites, for example
Twintex and GMT. Although shown to work on a laboratory scale, the technology
had not been industrialised until the development of the Phase 1 demonstrator
component tooling. The technique was implemented successfully, with a number of
mouldings being produced, although the work did highlight the need for tool
7-14

temperature to be considered, since heating of the tool during moulding caused


interference issues with the sprung shear edge. The blankholder had no such problems
and produced reliable results. Fabric wrinkling was not seen in any of the components
manufactured, in part due to the geometry of the component, but potentially also due
to the tension in the fibres during forming, applied by the blankholder.

7.5.2 The Effect of Process Variability


The only demonstrator component where manufacturing issues potentially
compromised performance was the 6mm Phase 2 demonstrator component. Test
results showed that the beam was significantly less stiff than expected, when tested in
3-point bending. After investigation of various potential explanations, the preheating
technique was identified as the cause. Both the 4mm thick beam and the 8mm thick
beam were manufactured from stacks of blanks preheated in units of four, containing
four 1mm thick layers of Twintex (one for the 4mm beam and two for the 8mm
beam). The 6mm beam was manufactured from a single stack of six 1mm layers,
preheated using the same technique as for the 4mm and 8mm beam. The measurement
of surface temperature both on the top and bottom of the stack meant that the internal
temperature had to be assumed to be high enough for moulding. The results for the
4mm and 8mm beams show that this was the case, but the result from the 6mm beams
suggest otherwise. The 6mm beams were moulded with a lower temperature in the
middle of the stack and therefore significantly underperformed.

It would be essential that if the process were to be industrialised, thorough moulding


trials were undertaken to optimise the processing parameters for each thickness of
beam. This would reduce variability and could also increase the performance of all the
beam geometries and thicknesses developed.

7.5.3 The Need for Part Integration


Thermoplastic composite components with polypropylene matrix materials, due to
their cost and mechanical properties need to be carefully designed to achieve
maximum benefit. This work has shown that, in principle, the concept of a
crashworthy Twintex part is a valid one, when considering performance alone. This
however is not the only criteria which an automotive manufacturer uses to assess
candidate materials. Composites, and in particular, non-exotic composites, need to
7-15

display a range of advantages before they are considered. One such advantage,
touched upon only briefly during this study and in general outside of the scope of the
research, is part integration.

Polypropylene is available in a range of forms, from the basic polymer through to


aligned fibre reinforced structural composite materials. These can all potentially be
combined in a single part and manufactured in a single moulding process. This is
therefore the area where not only the research engineer, but also the commercial
engineer, see their objectives converging. The potential to mould, in a single shot, a
pre-coloured door module, with trim, outer skin, inner structural reinforcement,
crashworthiness and lifetime corrosion resistance built in, is an attractive concept.
This is especially true since the biggest hurdle, that of matching the structural
performance of current steel designs, has been shown to be achievable in a relatively
short timescale and with relatively low investment in tooling and development costs.

7.6

Recommendations for Future Work

This research has covered a wide range of topics, including processing, modelling and
industrial component manufacture and design. It is proposed that the following areas
of further work be considered. Some of the proposals are minor changes to method or
approach, which have been highlighted when considering this work as a whole, whilst
others are suggestions of larger units of research, which could be undertaken to
further the knowledge and understanding of thermoplastic composites.

7.6.1 Inclusion of Residual load Carrying Capacity in the Implicit Model


Having assessed the results from the implicit finite element damage modelling work,
it is suggested that a residual strength be implemented in the model, in an attempt to
resolve stability issues associated with fibre direction damage in complex
components.

7.6.2 Implementation of the Implicit Damage Model in an Explicit Code


The implicit finite element damage model developed has been shown to account for
the predominant in-plane damage modes observed in Twintex and could therefore
potentially be implemented in an explicit finite element code. An explicit code could

7-16

deal with non-linearity and large displacement more effectively than the implicit
solution could, potentially allowing the model to be developed and improved.

7.6.3 Assessment of Alternative damage Models


The field of material modelling, in particular using explicit finite element codes, is
constantly evolving. It is therefore suggested that new models, becoming available
specifically for the characterisation of fabric composites, should be investigated.
Potentially these could be more suitable for composite materials with woven glass
fibre reinforcement, such as Twintex.

7.6.4 Assessment of Relevance of Delamination as a Damage Mechanism


It is felt that neglecting delamination as a damaging mode in thermoplastics, could
lead to incorrect prediction of performance. Although potentially not as critical for
thermoplastic matrix composites as it is for thermoset matrix composites the
phenomenon should be investigated, to confirm the validity of the current approach.

7.6.5 Rate Dependency


If Twintex is to be used in fully crashworthy structures, the material rate dependency
must be assessed. This was beyond the scope of the current work and not included in
the material models developed and investigated.

7.6.6 Development of a Fully Thermoplastic Door Concept


In terms of the industrial continuation of this work, it is felt that the next step is the
development of a fully thermoplastic door concept. The component parts, and in
particular the crashworthy structure, have been shown to work effectively. Therefore
the logical conclusion would be prototype development leading to maturity of the
technology into commercial applications.

7-17

7.7

Conclusions

Initially after a review of current literature, three areas were identified as requiring
further work. These have been investigated and the following conclusions are drawn.

Implicit finite element damage modelling techniques can be developed to account for
the damage and failure modes observed in a woven glass fibre reinforced
polypropylene composite. The de-coupling of shear and fibre direction damage can
result in a model that is applicable to a range of test scenarios. This approach can be
used as a design tool for more complex components, but is limited when considering
high levels of material non-linearity and damage development, due to the stability of
the implicit finite element method.

Current explicit finite element damage modelling techniques and in particular the biphase material model, implemented in PAM-CRASH, are applicable to aligned glass
fibre reinforced polypropylene composite materials, but with significant limitations,
especially when considering shear damaging behaviour. The application of this model
to a complex geometry has shown that the bi-phase model can be calibrated and
validated as a design tool for thermoplastic composite components, but only if
considering damage development that occurs predominantly in the fibre direction.

Finally, it is concluded that aligned glass fibre reinforced polypropylene composite


materials are suitable for structural automotive applications, such as side impact
protection and perform to a similar standard as steel components. If the weight
penalty observed at a component level is overcome by modularisation then this
technology could potentially be used commercially in crashworthy automotive
applications.

7-18

Chapter 8
[1]

References

Harrison, A., Driving Composites - Weighting For Improved Vehicle Fuel

Economy, 5th International Conference on Automated Composites - ICAC 97,


Glasgow, September 4-5, 1997, pp. 3-12.

[2]

Jambor, A. and Beyer, M., New Cars - New Materials, Materials & Design,

Vol. 18 (4/6), 1997, pp. 203-209.

[3]

McConnell, V. P. Robust Growth Predictions Outlined at SAE 2000,

Reinforced Plastics, May, 2000, pp. 34-38.

[4]

Quadrant Plastic Composites AG (Formerly Symalit AG) Thermoplastic

Composites

for

Your

Car

as

Well,

http://www.quadrantcomposites.com/

English/idqpc102.asp, Lenzburg, 2003.

[5]

Saint-Gobain Vetrotex, Twintex Markets and Applications - Transportation,

http://www.twintex.com/markets_app/tw_trans.html, France, 2003.

[6]

Cole, G. S. and Sherman, A. M., Lightweight Materials for Automotive

Applications, International Metallographic Society Symposium on Microstructural


Characterisaztion of Lightweight Materials for Transportation, Montreal, July 2425, 1994, pp. 3-9.

[7]

Svensson, N., Shishoo, R. and Gilchrist, M., Manufacturing of Thermoplastic

Composites from Commingled Yarns - A Review, Journal of Thermoplastic


Composite Materials, Vol. 11, 1998, pp.22-56.

[8]

Ye, L., Friedrich, K. and Kastel, J., Consolidation of GF/PP Commingled

Yarn Composites, Applied Composite Materials, Vol 1, 1995, pp. 415-429.

8-1

[9]

Klinkmuller, V., Um, M., Friedrich, K. and Kim, B., Impregnation and

Consolidation of Different GF/PP Co-mingled Yarn, Tenth International Conference


on Composite Materials, Whistler, Vol. 3, 1995, pp. 397-403.

[10]

Klinkmuller, V., Um, M. K., Steffens, M., Friedrich, K. and Kim, B. S., A

New Model for Impregnation Mechanics in Different GF/PP Commingled Yarns,


Applied Composite Materials, Vol. 1, 1995, pp. 351-371.

[11]

Wakeman, M., Cain, T., Rudd, C. D., Brooks, R. and Long, A. C.,

Compression Moulding of Glass and Polypropylene Composites for Optimised


Macro- and Micro-Mechanical Properties - 1 Commingled Glass and Polypropylene,
Composites Science and Technology, Vol 58, 1998, pp. 1879-1898.

[12]

Wakeman, M., Non-Isothermal Compression Moulding of Glass Fibre

Reinforced Polypropylene Composites, Ph.D. Thesis, University of Nottingham, UK,


1997.

[13]

Osten, S., St. John, C., Guillon, D., Zanella, G. and Renault, T., Compression

Molding of Twintex and Random Fiber Thermoplastic Molding Materials, ANTEC


97, The Annual Technical Conference of the Society of Plastics Engineers, Toronto,
1997, pp. 2432-2436.

[14]

Bruer, U. and Neitzel, M. High Speed Stamp Forming of Thermoplastic

Composite Sheets, Polymers and Polymer Composites, Vol. 4, No. 2, 1996, pp 117123.

[15]

Curtis, C., Energy Absorbtion and Crush Behaviour of Composite Tubes,

Ph.D. Thesis, University of Nottingham, UK, 2000.

[16]

Loureno, N. S. F., Predictive Finite Element Method for Axial Crush of

Composite Tubes, Ph.D. Thesis, University of Nottingham, UK, 2002.

[17]

Saint-Gobain

Vetrotex,

Physical

Properties

for

FEA

Modelling

TPEAT4460K, www.twintex.com/matprop/tw_physical.html, France, 2003.


8-2

[18]

Saint-Gobain Vetrotex, Data Base for 2-D Woven fabrics - DCC1, France,

1998.

[19]

Pierron, F. and Vautrin, A., Accurate Comparative Determination of the In-

Plane Shear Modulus of T300/914 by the Iosipescu and 45 Off-Axis Tests,


Composites Science and Technology, Vol. 52, 1994, pp. 61-72.

[20]

Symalit AG, GMT Data Sheet, Personal Communication, 1998.

[21]

Borealis A/S, PLYTRON GN638T Unidirectional Glass-Fibre/Polypropylene

Composite, Physical Properties, Denmark, 1997.

[22]

Gibson, R. F., Principles of Composite Material Mechanics, McGraw Hill

Inc., 1994.

[23]

Hull, D. and Cline, T. W., An Introduction to Composite Materials, 2nd

Edition, Cambridge University Press, 1996.

[24]

Cantwell, W. J. and Morton, J., The Impact Resistance of Composite

Materials - A Review, Composites, Vol. 22, No. 5, 1991, pp. 347-362.

[25]

Srinivasan, K., Jackson, W. C., Smith, B. T. and Hinkley, J. A.,

Characterization of Damage modes in Impacted Thermoset and Thermoplastic


Composites, Journal of Reinforced Plastics and Composites, Vol. 11, October 1992,
pp. 1111-1126.

[26]

Jouri, W. S. and Shortall, J. B., Impact Testing of Long Fibre reinforced

Thermoplastic Composites, Journal of Thermoplastic Composite Materials, Vol. 4,


July 1991, pp. 206-226.

[27]

Santulli, C., Brooks, R., Long, A. C., Warrior, N. A. and Rudd, C. D., Impact

Properties

of

Compression

Moulded

Commingled

E-glass-Polypropylene

Composites, Plastics, Rubber and Composites, Vol. 31, No. 6, 2002, pp. 270-277.
8-3

[28]

Ramakrishna, S. and Hamada, H., Energy Absorption Characteristics of

Crash Worthy Structural Composites, Key Engineering Materials, Vols. 141-143,


1998, pp. 585-620.

[29]

Sabic Europe, BMW Mini Front End Carrier, StaMax Commercial

Application Datasheet, Dec, 2001.

[30]

Anon, Mini Cooper has StaMax Front-End Carrier, http://www.sae.org/

automag/material/12-2001/, Dec, 2001.

[31]

Anon,

OC

Showcases

SUV

Applications

at

SAE

2003,

http://www.reinforcedplastics.com/WZ/RPlastics/market_focus/automotive/000017/
show/, 2003.

[32]

Sabic Europe, Ford Fiesta Door Modules, StaMax Commercial Application

Datasheet, Dec, 2001.

[33]

Bricout, A., Composite materials: Performance of Twintex for Automotive

Weight Saving, www.twintex.com, 1998.

[34]

Hexcel Composites, TowFlex Fabric Products Provide Performance and

Manufacturing Benefits for BMW M3 Bumper Beams, TowFlex thermoplastics


Applications,

http://www.hexcelcomposites.com/Markets/Products/Thermoplastics/

Applications/Bumper, 2001.

[35]

Anon, Hexcel Thermoplastic TowFlex featured in BMW M3 Bumper,

http://composite.miningco.com/library/PR/2001/blhexcel21.htm, 2001.

[36]

Warrior, N., Polymer Composites for Automotive Crashworthiness,

Automotive Composites, Warwick University IMC, Coventry, June 10, 2003.

[37]

Pitrof, S. M. and Merrifield, R. A., Engineering Development of a Composite

Cross Car Beam, SAE Technical Paper 970727, 1997.


8-4

[38]

Manwaring, D., Chan, A. and Marks, M., A Structural Instrument Panel from

Glass-Mat thermoplastic for the Small-Car Market , SAE Technical Paper 970726,
1997.

[39]

Anon, Plastic/Metal Hybrid Composite technology IP Support Beam, Bayer

Corporation,

CLN

Case

Study,

http://www.lightspeed.com/

csstdies/

links/ipbeam.html, 2000.

[40]

Anon, Composite B-Post Passes the Test, European Automotive Design,

March 199, pp 11.

[41]

Sherman, K. C. and Florence, R., Design and Development of an Engineering

Thermoplastic Energy Absorbing System for Automotive Knee Bolsters , SAE


Technical Paper 970725, 1997.

[42]

Davis, R., Ford Motor Company, Personal Communication, 24 Feb 1999.

[43]

European Commission, ELV - End of Life Vehicle Directive, Proposal

COM(97)358

amended

COM(99)1761997,

http://europa.eu.int/comm/

environment/docum/index.html.

[44]

EURO NCAP, EURO NCAP Side Impact - How the Tests are Done,

http://www.euroncap.com/tests.htm, 2003.

[45]

National Highway Traffic Safety Administration, Federal Motor Vehicle

Safety Standard (49 CFR Part 571) MVSS214 Side Impact Protection, Federal
Register, Vol. 58, No. 225, Issue 3, December 1995.

[46]

Lie, A. and Tingvall, C., How Does Euro NCAP Results Correlate to real

Life Injury Risks - A Paired Comparison Study of Car-to-Car Crashes, IRCOBI,


International Conference on the Biomechanics of Impact, Montpelier, France, Sept
20, 2000.

8-5

[47]

Kahane, C., An Evaluation of Side Structure Improvements in Response to

Federal Motor Vehicle Safety Standard 214, NHTSA Report Number DOT HS 806
314, Nov 1982.

[48]

Ray, M. H., Impact Conditions in Side-Impact Collisions with Fixed

Roadside Objects, Accident Analysis and Prevention, Vol. 31, 1999, pp. 21-30.

[49]

Miltner, E. and Salwender, H. J., Injury Severity of Restrained Front Seat

Occupants in Car-to-Car Side Impacts, Accident Analysis and Prevention, Vol. 27,
No. 1, 1995, pp. 105-110.

[50]

Farmer, C. M., Braver, E. R. and Mitter, E. L., Two-Vehicle Side Impact

Crashes: The Relationship of Vehicle and Crash Characteristics to Injury Severity,


Accident Analysis and Prevention, Vol. 29, No. 3, 1997, pp. 399-406.

[51]

Mizuno, K. and Kajzer, J., Compatibility Problems in Frontal, Side, Single

Car Collisions and Car-to-Pedestrian Accidents in Japan, Accident Analysis and


Prevention, Vol. 31, 1999, pp. 381-391.

[52]

Uduma, K., Innovations in Auto Safety design, a Key to Quality

Improvement, Technological Forecasting and Social Change, Vol. 64, 2000, pp. 197208.

[53]

Koehr,

R.,

ULSAC

Lighweight

Steel

Automotive

Closures,

AutoTechnology, Vol. 1, 2001, pp. 74-77.

[54]

Cheon, S. K., Lee, D. G. and Jeong, K. S., Composite Side-Door Impact

Beams for Passenger Cars, Composite Structures, Vol. 38, No. 1-4, 1997, pp. 229239.

[55]

Kamil, R. and Saunders, E., Side Impact Beam Design, MAE226 -

Mechanics

of

Composite

Materials,

Project

Report,

http://spline.mae.wvu.edu/Courses6/projects_S96/ERIC/sidebeam.html.

8-6

2000,

[56]

Adam, H., Patberg, L., Philipps, M. and Dittmann, R., Testing of New

composite Side Door Concepts, SAE Special Publications, 980859, Vol. 1320, Feb
1998, pp. 23-30.

[57]

Erzen, S., Ren, Z. and Anzel, I., Analysis of FRP Side-Door Impact Beam,

Faculty of Mechanical Engineering, University of Maribor, Slovenia, 2002,


http://www.sussex.ac.uk/automotive/tut2002/15_erzen.pdf.

[58]

Clemo, K., SACTAC Project: Door Beam Design Sub-group, Progress

Reports 1-5, MIRA/University of Nottingham, Dec 1996-Jan 1998.

[59]

Clemo, K. and Price, C., Developing Composite Material Side-Impact

Beams, MIRA New Technology, 2000, pp. 184-190.

[60]

Priston, A-M., Evaluation of Stress Induced Damage in Composite Material,

PhD Thesis, Nottingham Trent University, UK, 1997.

[61]

Cheikh, M., Reanalysis of Structures with Modification of the Stiffness,

Application to the Modelling of Damage in Unidirectional Composites, Composites


and Structures, Vol. 78, 2000, pp. 725-736.

[62]

Belingardi, G., Gugliotta, A. and Vadori, R., Numerical Simulation of

fragmentation of Composite Material Plates due to Impact International journal of


Impact Engineering, Vol. 21, No. 5, 1998, pp. 335-347.

[63]

Feng, Z.-N., Allen, H. G. and Moy, S. S. J., Theoretical and Experimental

Investigation of Progressive Failure of Woven Composite Panels, Journal of


Composite Materials, Vol. 33, No. 11, 1999, pp. 1030-1047.

[64]

Gamble, K., Pilling, M. and Wilson, A., An Automated Finite Element

Analysis of the Initiation and Growth of Damage in Carbon Fibre Composite


Materials, Composite Structures, Vol. 32, 1995, pp. 265-274.

8-7

[65]

Tan, S. C. and Perez, J., Progressive Failure of Laminated Composites with a

Hole Under Compressive Loading, Journal of Reinforced Plastics and Composites,


Vol. 12, 1993, pp. 1043-1056.

[66]

Padhi, G. S., Shenoi, R. A., Moy, S. S. J. and Hawkins, G. L., Progressive

Failure and Ultimate Collapse of Laminated Composite Plates in Bending,


Composite Structures, Vol. 40, Nos. 3-4, 1998, pp. 277-291.

[67]

Ozden, O. O. and Engblom, J. J., Analysis of Progressive Failure in

Composites, Composites Science and Technology, Vol. 28, 1987, pp. 87-102.

[68]

Chang, F. and Chang, K., A Progressive Damage Model for Laminated

Composites Containing Stress Concentrations, Journal of Composite Materials, Vol.


21, Sept 1987, pp. 834-855.

[69]

Chow, C. L. and Yang, F., Inelastic Finite Element Analysis of Fibre-

Reinforced Composite Laminates with Damage, Proceedings of the IMechE, Vol.


212, Part C, pp. 717-729.

[70]

Williams, K. V., Floyd, A. M., Vasiri, R. and Poursartip, A., Numerical

Simulation of In-Plane Damage Progression in Laminated Composite Plates, 12th


International Conference on Composite Materials - ICCM 12, Paris, 1999.

[71]

Vang, L., Baubakar, M. L., Trivaudey, F. and Perreux, D., Coupled Damage

Elasto-Dissipative Models for Numerical Analysis of Laminate Shells, 13th


International Conference on Composite Materials - ICCM 13, Beijing, 2000.

[72]

Oytana, C., Damage Indicators Part I Mechanics and Mechanisms of

damage in Composites and Multi-Materials, ESES11 (Edited by D. Baptiste), 1991,


Mechanical Engineering Publications, London, pp. 215-232.

8-8

[73]

Coats, T. W., Harris, C. E., Lo, D. C. and Allen, D. H., Progressive damage

Analysis of Laminated Composite (PDALC) - (A Computational Model Implemented


in the NASA COMET Finite Element Code) Version 2.0, NASA Technical Report
NASA/TM-1998-208718, 1998.

[74]

Coats, T. W. and Harris, C. E., A Damage-Dependant Finite Element

Analysis for Fiber-Reinforced Composite Laminates, NASA Technical Paper, 1998.

[75]

Harris, C. E., Coats, T. W. and Glaessgen, E. H., Experimental Verification

of Computational Models for Laminated Composites, NASA Technical Paper, 1998.

[76]

Iannucci, L., Failure Propagation Modelling Using Explicit Codes NAFEMS

Awareness Seminar Advances in Composites Analysis & Design, London, 20th June
2001.

[77]

Iannucci, L. and Willows, M. J., Progressive Failure Modelling of Woven

Carbon Composites Under Impact, Euromech 40, European Mechanics Society


Colloquium, Imperial College, London, 27-29 Sept 1999.

[78]

Iannucci, L., Dechaene, R., Willows, M. J. and Degrieck, J., A Failure Model

for the Analysis of Thin Woven Glass Composite Structures Under Impact Loadings,
Composites and Structures, Vol. 79, 2001, pp. 785-799.

[79]

Chang, F. and Lessard, L. B., Damage Tolerance of Laminated Composite

Plates Containing an Open Hole and Subjected to Compressive Loadings: Part I Analysis, Journal of Composite Materials, Vol. 25, Jan 1991, pp. 2-43.

[80]

Ladeveze, P., On a Damage Mechanics Approach, Mechanics and

Mechanisms of damage in Composites and Multi-Materials, ESES11 (Edited by D.


Baptiste), 1991, Mechanical Engineering Publications, London, pp. 119-141.

[81]

E.S.I. Group, PAM-CRASH/PAM-SAFE Solver Notes Manual, Version

1998, France, 1998.

8-9

[82]

Hahn, H. T. and Tsai, S. W., Nonlinear Elastic Behaviour of Unidirectional

Composite Laminates, Journal of Composite Materials, Vol. 7, 1973, pp. 102-110.

[83]

Lessard, L. B. and Chang, F., Damage Tolerance of Laminated Composite

Plates Containing an Open Hole and Subjected to Compressive Loadings: Part II Experiment, Journal of Composite Materials, Vol. 25, Jan 1991, pp. 44-64.

[84]

Chang, K., Liu, S. and Chang, F., Damage Tolerance of Laminated

Composites Containing an Open Hole and Subjected to tensile Loadings, Journal of


Composite Materials, Vol. 25, March 1991, pp. 274-301.

[85]

Avalle, M., Belingardi, G. and Vadori, R., Numerical Simulation of Impact

Damage in Composite Material Structures, Impact and Dynamic Fracture of


Polymers and composites, ESES19 (Edited by J. G. Williams and A. Pavan), 1995,
Mechanical Engineering Publications, London, pp. 329-340.

[86]

Shahid, I. and Chang, F., An Accumulative Damage Model for tensile and

Shear Failure of Laminated Composite Plates, Journal of Composite Materials, Vol.


29, No. 7, 1995, pp. 274-301.

[87]

Davila, C. G., Ambur, D. R. and McGowan, D. M., Analytical Prediction of

Damage Growth in Notched Composite Panels Loaded in Axial Compression, Paper


AIAA-99-1435, 40th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics
and Materials Conference, St. Louis, April 12-15, 1999.

[88]

Hibbitt, Karlsson and Sorenson Inc. ABAQUS Examples Manual, USA,

1998.

[89]

Ladeveze, P., A Damage Computational Method for Composite Structures,

Computers and Structures, Vol. 44, No. 1/2, 1992, pp. 79-87.

[90]

Ladeveze, P., A Damage Computational Approach for Composites: Basic

Aspects and Micromechanical Relations, Computational Mechanics, Vol. 17, 1995,


pp. 142-150.
8-10

[91]

Ladeveze, P. and le Dantec, E., Damage Modelling of the Elementary Ply for

Laminated Composites, Composites Science and technology, Vol. 43, 1992, pp. 257267.

[92]

Allix, O., Daudeville, L. and Ladeveze, P., Delamination and Damage

Mechanics, Mechanics and Mechanisms of damage in Composites and MultiMaterials, ESES11 (Edited by D. Baptiste), 1991, Mechanical Engineering
Publications, London, pp. 143-157.

[93]

Allix, O., Ladeveze, P. and Corigliano, A., Damage Analysis of Interlaminar

Fracture Specimens, Composite Structures, Vol. 31, 1995, pp. 61-74.

[94]

Allix, O., Guedra-Degeorges, D., Guinard, S. and Vinet, A., 3D Damage

Analysis Applied to Low-Energy Impacts on Composite Laminates, 12th


International Conference on Composite Materials - ICCM 12, Paris, 1999.

[95]

Ladeveze, P., Allix, O., Deu, J. and Leveque, D., A Mesomodel for

Localisation and Damage Computation in Laminates, Computer Methods in Applied


Mechanics and Engineering, Vol. 183, 2000, pp. 105-122.

[96]

Touchard, F., Lafarie-Frenot, M. C. and Guedra-Degeorges, D., Mechanical

Behaviour Characteristics of a Thermoplastic Composite Used in Structural


Components, Composites Science and Technology, Vol. 56, 1996, pp. 785-791.

[97]

Coutellier, D. and Rozycki, P., Multi-Layered Multi-Material Finite Element

for crashworthiness Studies, Composites: Part A, Vol. 31, 2000, pp. 841-851.

[98]

Hochard, C., Auborg, P.-A. and Charles, J.-P., Modelling of the Mechanical

Behaviour of Woven-Fabric CFRP Laminates up to Failure, Composites Science and


technology, Vol. 61, 2001, pp. 221-230.

8-11

[99]

Johnson, A. F. and Simon, J., Modelling Fabric Reinforced Composites

Under Impact Loads, Euromech 400, European Mechanics Society Colloquium,


Imperial College, London, 27-29 Sept 1999.

[100] de Rouvray, A. and Haug, E., Failure of Brittle and Composite Materials by
Numerical Methods, Chapter 7 in Structural Failure, Editor T. Wierzbicki, John
Wiley & Sons, 1989.

[101] de Rouvray, A., Haug, E. and Stavrinidis, C., Composite material damage
and fracture Models for Numerical Simulations in Support of Fracture Control,
International Conference on Spacecraft Structures and Mechanical testing,
Noordwijk, April 24-26, 1991.

[102] Pickett, A., Ruckert, J., Ulrich, D. and Haug, E., Material damage Law
Suitable for Crashworthiness Investigation of Random Fibre Composite Materials,
XVIII International Finite Element Congress, Baden-Baden, Nov 20-21, 1989.

[103] de Rouvray, A., Dowlatyari, P. and Haug, E., Validation of the PAMMFISS/Bi-Phase Numerical Model for the Damage and Strength Prediction of LFRP
Composite Laminates, Mechanics and Mechanisms of Damage in Composite and
Multi-Materials, ESIS11, Editor D. Baptiste, Mechanical Engineering Publications,
London, 1991, pp. 183-202.
[104] Haug, E. and de Rouvray, A., Crash response of Composite Structures, 3rd
International Symposium on Structural Crashworthiness, Liverpool, April 14-16,
1993, pp. 237-294.

[105] Haug, E. and Jamjian, M., Numerical Simulation of the Impact Resistance of
Composite Structures, Chapter 8 in Analysis and Modelling of Composite
Materials, 1996.

[106] Haug, E. and Jamjain, M., Industrial Crashworthiness Simulation of


Automotive Structures and Components Made of Continuous Fiber Reinforced
Composite and Sandwich Assemblies, ESI SA Paper No. 94UL026, France, 1994.
8-12

[107] Haug, E., Watanabe, M. and Nakada, I., Numerical Crashworthiness


Simulation of Automotive Structures and Components Made of Continuous Fiber
Reinforced Composite and Sandwich Assemblies, SAE International Congress and
Exhibition, Detroit, Feb, 1991.

[108] Nakada, I. and Haug, E., Numerical Simulation of Crash Behaviour of


Composite Structures for Automotive Applications, TechMat 92 - Euro-Japan
Exchanges on Materials, Paris, 1992.

[109] Johnson, A. F., Kindervater, C. M., Kohlgruber, D. and Lutzenburger, M.,


Predictive Methodologies for the Crashworthiness of Aircraft Structures, American
helicopter Society 52nd Annual Forum, Washington D.C., June 4-6, 1996.

[110] Kohlgruber, D. and Kamoulakos, A., Validation of Numerical Simulation of


Composite Helicopter Sub-Floor Structures Under Crash Loading, American
helicopter Society 54th Annual Forum, Washington D.C., May 20-22, 1998.

[111] Deletombe, E., Delsart, D., Kohlgruber, D. and Johnson, A. F., Improvement
of Numerical methods for Crash Analysis in Future Composite Aircraft design,
Aerospace Science and Technology, Vol. 4, 2000, pp. 189-199.

[112] McCarthy, M. A., Harte, C. G., Wiggenraad, J. F. M., Michielsen A. L. P. J.,


Kohlgruber, D. and Kamoulakos, A., Finite Element Modelling of Crash response of
Composite Aerospace Sub-Floor structures, Computational Mechanics, Vol. 26,
2000, pp. 250-258.

[113] McCarthy, M. A. and Wiggenraad, J. F. M., Numerical Investigation of a


Crash Test of a Composite Helicopter Subfloor Structure, Composite Structures,
Vol. 51, 2000, pp. 345-359.

[114] BS EN ISO 527-4, Plastics - Determination of tensile properties - Part 4: Test


conditions for isotropic and orthotropic fibre-reinforced composites, 1997

8-13

[115] ASTM D 3410 75, Standard Test Method for Compressive Properties of
Oriented Fiber Composites, 1975

[116] Wan, T. and Tham, L. Data for Modelling the Impact of Automotive
Laminates, BEng Final Year Project Thesis, University of Nottingham, UK, 2002.

[117] Duckett, M. J., Rate Dependent Effects on the Energy Absorption and
Material Properties of Polymer Composites, PhD Thesis, University of Nottingham,
UK, October 2001.

[118] Hibbitt, Karlsson and Sorenson Inc. ABAQUS/Standard Users Manual Volume I, USA, 1998.

[119] Thomas, A. V., SACTAC Project: Modular Door Cassette Design Sub-group,
Progress Reports 1 & 2, Jaguar Cars Limited, March 1996-June 1996.

[120] EN ISO 14129, Fibre-reinforced Plastic Composites - Determination of Inplane Shear Stress/Shear Strain Response, Including the In-plane Shear Modulus and
Strength, by the 45 Tension Test Method, 1997.

[121] Santulli, C., CRACTAC Project Progress Update, Internal Communication,


University of Nottingham, 2000.

[122] Pickett, A., "Design and Manufacture of Textile Composites", Ed. A. C. Long,
Ch. 8, Woodhead Publishing Ltd, (in press).

8-14

Appendix A

ABAQUS/Standard User Defined Field

The following source code, adapted from a previously implemented algorithm [88],
was used to implement the damage model proposed in Chapter 4 and calculate the
field variables for an ABAQUS/Standard user defined material.

** --------------------** USER-DEFINED FIELD


** --------------------*USER SUBROUTINES
SUBROUTINE USDFLD(FIELD,STATEV,PNEWDT,DIRECT,T,CELENT,TIME,DTIME,
1 CMNAME,ORNAME,NFIELD,NSTATV,NOEL,NPT,LAYER,KSPT,KSTEP,KINC,
2 NDI,NSHR)
C
INCLUDE 'ABA_PARAM.INC'
C
C SHEAR MODULUS AND DAMAGE PARAMETER
PARAMETER(G12=1.04D3,ALPHA=1.4D-5)
C
CHARACTER*80 CMNAME,ORNAME
CHARACTER*8 FLGRAY(15)
DIMENSION FIELD(NFIELD),STATEV(NSTATV),DIRECT(3,3),T(3,3),TIME(2)
DIMENSION ARRAY(15),JARRAY(15)
C
C INITIALIZE FAILURE FLAGS FROM STATEV.
EFF
= STATEV(1)
EFS
= STATEV(2)
DAMAGE = STATEV(3)
C
C GET STRESSES FROM PREVIOUS INCREMENT
CALL GETVRM('S',ARRAY,JARRAY,FLGRAY,JRCD)
S11 = ARRAY(1)
S22 = ARRAY(2)
S12 = ARRAY(4)
C
C GET SHEAR STRAIN FROM PREVIOUS INCREMENT
CALL GETVRM('E',ARRAY,JARRAY,FLGRAY,JRCD)
E12 = ARRAY(4)
C
C SHEAR DAMAGE INDEX: = 0 IF NO STRAIN TO PREVENT DIVIDE BY ZERO
C
IF (E12.NE.0) THEN
DAMAGE = (3.D0*ALPHA*G12*S12**2 - 2.D0*ALPHA*(S12**3)/E12) /
&
(1.D0 + 3.D0*ALPHA*G12*S12**2)
ELSE
DAMAGE = 0.D0
ENDIF
C
C
C PLY TENSILE/COMPRESSIVE FAILURE
C
IF (EFF .LT. 1.D0) THEN
IF (S22 .LT. -137) THEN
EFF=2
END IF
IF (S22 .GT. 279) THEN
EFF=2
END IF
IF (S11 .LT. -137) THEN
EFF=2
END IF
IF (S11 .GT. 279) THEN
EFF=2
END IF
STATEV(1) = EFF
ENDIF

A-1

C
C PLY ULTIMATE SHEAR FAILURE
C
IF (EFS .LT. 1.D0) THEN
IF (E12 .GT. 0.45) THEN
EFS=2
ELSE IF (E12 .LT. -0.45) THEN
EFS=2
ELSE
EFS=0
ENDIF
STATEV(2) = EFS
ENDIF
C
C
UPDATE FIELD VARIABLES
C
FIELD(1) = 0.D0
FIELD(2) = 0.D0
IF (EFF .GT. 1.D0) FIELD(1) = 1.D0
IF (EFS .GT. 1.D0) FIELD(2) = 1.D0
FIELD(3) = DAMAGE
STATEV(3) = FIELD(3)
C
RETURN
END

A-2

Appendix B

Derivation of Shear Damage Formulation

The shear damage model implemented in the ABAQUS/Standard user defined field
given in Appendix A, was based on the model proposed by Chang and Lessard [79].
The algorithm used was adapted to improve stability during analysis. This adaptation
was developed from a previous implementation of the model [88] and is included here
for completeness.

The original form of the model,

12 =

12
G12

+ 123

(B.1)

can be rearranged to:

12 = G121 12 + 123

(B.2)

For an increment, i, the shear stress can then be expressed as a linear function of
strain,

12(i +1) = G121 + ( 12(i ) ) 12(i +1)

(B.3)

G12
1 + G (i )
12
12

(i +1)

2 12

(B.4)

which is inverted to:

12(i +1) =

( )

This gives an algorithm which will allow the definition of effective shear modulus
over an increment, i.

This algorithm is not stable at higher strain levels [88]. This can be demonstrated by
considering an increment where strain is constant, such that:

12(i +1) = 12(i ) = 12

(B.5)

If the stress at increment i is considered to have a small perturbation from the exact
e (i )
solution at that increment 12
, then the solution at that increment is,
(i )

12(i ) = 12e + 12(i )

A-3

(B.6)

and similarly at an increment i+1 is:

12(i +1) = 12e


For the algorithm to be stable

12(i +1)

(i +1)

12
where

12 = 12e

12(i )

should be no larger than

The perturbation at i+1 is calculated by substituting


e (i )
linearising about 12
:
(i +1)

(B.7)

+ 12(i +1)

12(i )

into equation B.4 and

2G122 12 12
=
12(i )
2
1 + G12 12

(B.8)

(i )

The perturbation at i+1 is larger than at i if:

2G122 1212 > 1 + G12 122

(B.9)

Eliminating 12 from equation B.9 gives.

123 > G121 12

(B.10)

Instability occurs at strain levels where the non-linear part of the shear strain is larger
than the linear part. This would significantly reduce the effectiveness of the algorithm
in the current study.

To obtain a more stable algorithm the nonlinear stress/strain law is written including a
coefficient [88].

12 + 123 = G121 12 + ( + ) 123

(B.11)

Equation B.11 can be linearised to the form,

(1 + (

(i ) 3
12

( ))
2

/ 12(i ) 12(i +1) = G121 + ( + ) 12(i ) 12(i +1)

A-4

(B.12)

which when inverted gives:


3

1 + 12(i ) / 12(i )

( )

(i +1)

12 =

1 + ( + )G12 12(i )

( )

G1212(i +1)

(B.13)

Following the same procedure as the original algorithm it can be shown that a small
(i )
perturbation 12 in increment i, is reduced to zero in i+1 if = 2 .
The stable algorithm can be expressed as:
3

(i +1)

12 =

1 + 2 12(i ) / 12(i )

( )
1 + 3G ( ( ) )
12

G1212(i +1)

i 2
12

(B.14)

which can be rearranged into the form:

12(i +1) = (1 d s )G1212(i +1)

(B.15)

where the shear damage level ds is defined as:


2

ds =

3G12 ( 12(i ) ) 2 ( 12(i ) ) 12(i )


2

1 + 3G12 ( 12(i ) )

(B.16)

It is in this form, in equation B.15, that the model is implemented in the user defined
field FORTRAN subroutine.

A-5

Appendix C

ABAQUS/Standard Material Cards

The following ABAQUS/Standard elastic laminate material control cards, with three
field dependencies, were calibrated from experimental test data, for balanced weave,
60% weight fraction glass reinforced polypropylene Twintex.

**
** MATERIAL: NONLINEAR SHEAR WITH BUILT-IN EXPLICIT FAILURE
**
** FV1: PLY COMPRESSIVE/TENSILE FAILURE
** FV2: PLY SHEAR FAILURE
** FV3: SHEAR DAMAGE (NONLINEARITY) PRIOR TO FAILURE
** TOTAL OF 2^3 = 8 STATES
**
*MATERIAL, NAME=TWIN11
*ELASTIC,
TYPE=LAMINA,
DEPENDENCIES=3
12.17E3,
12.7E3,
0.08, 1.04E3,
1.52E3,
1.52E3,
0.,
1.00,
1.00,
0.00, 1.04E3,
1.52E3,
1.52E3,
0.,
12.17E3,
12.7E3,
0.00,
1,
1.52E3,
1.52E3,
0.,
1.00,
1.00,
0.00,
1,
1.52E3,
1.52E3,
0.,
12.17E3,
12.7E3,
0.08,
25,
1.52E3,
1.52E3,
0.,
1.00,
1.00,
0.00,
25,
1.52E3,
1.52E3,
0.,
12.17E3,
12.7E3,
0.00,
1,
1.52E3,
1.52E3,
0.,
1.00,
1.00,
0.00,
1,
1.52E3,
1.52E3,
0.,
*DEPVAR
3
**

A-6

0.,
1,
0.,
1,
0.,
1,
0.,
1,

0.,
0.,
1,
1,
0.,
0.,
1,
1,

0.
0.
0.
0.
1
1
1
1

Appendix D

PAM-CRASH Material Cards

The following PAM-CRASH degenerate bi-phase material control cards were


calibrated from experimental test data, for balanced weave, 60% weight fraction glass
reinforced polypropylene Twintex.

$
$ MATERIAL DATA CARDS
$
$---5---10----5---20----5---30----5---40----5---50----5---60----5---70----5---80
MATER /
1
130
1.5e-06
1
Ply
0
0
0.01
0.01
0.01
0.8333
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.800
40
0
0
1
1
1
2
1
3
1
4
1
5
1
6
1
7
1
8
1
9
1
10
1
11
1
12
1
13
1
14
1
15
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
$---5---10----5---20----5---30----5---40----5---50----5---60----5---70----5---80
PLY
/
1
0
0
0
Twintex 1:1 0/90
12.70
12.70
5.30
1.040
1.520
1.520
0.080
0.080
0.360
0.0053
0.017
0.039
0.25
0.9
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
1
0
0.5
11.40
11.40
5.30
1.040
1.520
1.520
0.080
0.080
0.360
0.0053
0.017
0.039
0.50
0.9
0
0
0
0
0
0
0
0
0
0
0
0
$
1
0
0

A-7

Appendix E
[i]

CRACTAC Project Publications

Rudd, C. D., Brooks, R., Long, A. C., Warrior, N. A., Wilson, M. J. and

Santulli, C., Development of a Crashworthy Composite Side Intrusion Beam Using


Long Glass Fibre Reinforced Polypropylene for the Foresight Vehicle Programme,
SAE2002, Detroit, 2002.

[ii]

Santulli, C., Brooks, R., Long, A. C., Warrior, N. A. and Rudd, C. D., Impact

Properties

of

Compression

Moulded

Commingled

E-glass-Polypropylene

Composites, Plastics, Rubber and Composites, Vol. 31, No. 6, 2002, pp. 270-277.

[iii]

Warrior, N. A., Rudd, C. D., Brooks, R., Long, A. C., Wilson, M. J. and

Santulli, C., Development Of A Thermoplastic Composite Side Intrusion Beam,


Materials for Lean Weight Vehicles 4, Motor Heritage Centre, Gaydon, October 3031, 2001.

[iv]

Warrior, N. A., Wilson, M. J., Brooks, R. and Rudd, C. D., Modelling of

Glass Reinforced Thermoplastic Composite Side Impact Structures, International


Journal of Crashworthiness, Vol. 6, No. 4, 2001.

[v]

Santulli, C., Brooks, R., Long, A. C., Rudd, C. D., Wilson, M. J. and Warrior,

N. A., Impact Properties of Thermoplastic Laminates for Automotive Applications,


Automotive Composites and Plastics ACP 2000, Ford Motor Company, Dunton, 5-6
December, 2000.

[vi]

Wilson, M. J., Warrior, N. A., Brooks, R. and Rudd C. D., Post First-Ply

Failure Modelling of Glass Reinforced Thermoplastic Composites, FRC 2000 Composites for the Millennium, Newcastle, September 13-15, 2000, pp. 505-511.

[vii]

Warrior, N. A., Wilson, M. J., Brooks, R. and Rudd, C. D., Modelling of

Glass Reinforced Thermoplastic Composite Side Impact Structures, ICrash 2000,


London, September, 2000.

A-8

S-ar putea să vă placă și